{
  "dataset": "tensor.news verified model scores",
  "description": "Every published model score on tensor.news with its metric, its evidence status, the evaluator that produced it, and the integrity grade of the benchmark it belongs to.",
  "source": "https://tensor.news/data",
  "updated_at": "2026-11-18T00:07:54.483368+00:00",
  "license": "CC BY 4.0 — attribution required: cite https://tensor.news/data",
  "count": 2995,
  "rows": [
    {
      "model": "DeepSeek-V3.2",
      "model_slug": "deepseek-v3-2",
      "developer": "DeepSeek",
      "benchmark": "ARC-AGI-2",
      "score": 4.03,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V3.2",
      "model_slug": "deepseek-v3-2",
      "developer": "DeepSeek",
      "benchmark": "ProofBench",
      "score": 8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2025-12-01",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3.2",
      "model_slug": "deepseek-v3-2",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 81.14,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-01",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3.2",
      "model_slug": "deepseek-v3-2",
      "developer": "DeepSeek",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 87.81,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-01",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3.2",
      "model_slug": "deepseek-v3-2",
      "developer": "DeepSeek",
      "benchmark": "DTBench",
      "score": 76,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-12-01",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V3.2",
      "model_slug": "deepseek-v3-2",
      "developer": "DeepSeek",
      "benchmark": "LMCA",
      "score": 33.91,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-12-01",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V3.2",
      "model_slug": "deepseek-v3-2",
      "developer": "DeepSeek",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 38.77,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-01",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V3.2",
      "model_slug": "deepseek-v3-2",
      "developer": "DeepSeek",
      "benchmark": "Aider polyglot",
      "score": 74.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-12-01",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3.2",
      "model_slug": "deepseek-v3-2",
      "developer": "DeepSeek",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 3.5,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-01",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3.2",
      "model_slug": "deepseek-v3-2",
      "developer": "DeepSeek",
      "benchmark": "Terminal Bench",
      "score": 39.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2025-12-01",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3.2",
      "model_slug": "deepseek-v3-2",
      "developer": "DeepSeek",
      "benchmark": "CL-bench",
      "score": 12.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-01",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V3.2",
      "model_slug": "deepseek-v3-2",
      "developer": "DeepSeek",
      "benchmark": "CL-bench Life",
      "score": 7.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-01",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3.2",
      "model_slug": "deepseek-v3-2",
      "developer": "DeepSeek",
      "benchmark": "ARC-AGI",
      "score": 57,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-01",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V3.2",
      "model_slug": "deepseek-v3-2",
      "developer": "DeepSeek",
      "benchmark": "APEX-Agents",
      "score": 21.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-01",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "gpt-oss-20b",
      "model_slug": "gpt-oss-20b",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 49.72,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "gpt-oss-20b",
      "model_slug": "gpt-oss-20b",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 65.24,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "gpt-oss-20b",
      "model_slug": "gpt-oss-20b",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 46.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-08-05",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "gpt-oss-20b",
      "model_slug": "gpt-oss-20b",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 40.93,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-08-05",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "gpt-oss-20b",
      "model_slug": "gpt-oss-20b",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 17.07,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-08-05",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "gpt-oss-20b",
      "model_slug": "gpt-oss-20b",
      "developer": "OpenAI",
      "benchmark": "Terminal Bench",
      "score": 3.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Terminal-Bench v2 Leaderboard",
      "measured_at": "2025-08-05",
      "source": "Terminal-Bench v2 Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "gpt-oss-120b",
      "model_slug": "gpt-oss-120b",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 70.5,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "gpt-oss-120b",
      "model_slug": "gpt-oss-120b",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 88.88,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "gpt-oss-120b",
      "model_slug": "gpt-oss-120b",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 60.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-08-05",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "gpt-oss-120b",
      "model_slug": "gpt-oss-120b",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 48.17,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "WeirdML Leaderboard",
      "measured_at": "2025-08-05",
      "source": "WeirdML Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "gpt-oss-120b",
      "model_slug": "gpt-oss-120b",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 26.05,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-08-05",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "gpt-oss-120b",
      "model_slug": "gpt-oss-120b",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 6.52,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2025-08-05",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "gpt-oss-120b",
      "model_slug": "gpt-oss-120b",
      "developer": "OpenAI",
      "benchmark": "METR Time Horizons",
      "score": 56.63,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-08-05",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "gpt-oss-120b",
      "model_slug": "gpt-oss-120b",
      "developer": "OpenAI",
      "benchmark": "Aider polyglot",
      "score": 41.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Aider LLM Leaderboards",
      "measured_at": "2025-08-05",
      "source": "Aider LLM Leaderboards",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "gpt-oss-120b",
      "model_slug": "gpt-oss-120b",
      "developer": "OpenAI",
      "benchmark": "Fiction.LiveBench",
      "score": 44.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-08-05",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "gpt-oss-120b",
      "model_slug": "gpt-oss-120b",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "gpt-oss-120b",
      "model_slug": "gpt-oss-120b",
      "developer": "OpenAI",
      "benchmark": "Surface Evolver Bench",
      "score": 25,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2025-08-05",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "gpt-oss-120b",
      "model_slug": "gpt-oss-120b",
      "developer": "OpenAI",
      "benchmark": "Terminal Bench",
      "score": 18.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Terminal-Bench v2 Leaderboard",
      "measured_at": "2025-08-05",
      "source": "Terminal-Bench v2 Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "gpt-oss-120b",
      "model_slug": "gpt-oss-120b",
      "developer": "OpenAI",
      "benchmark": "APEX-Agents",
      "score": 4.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "gpt-oss-120b",
      "model_slug": "gpt-oss-120b",
      "developer": "OpenAI",
      "benchmark": "Lech Mazur Writing",
      "score": 77.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-08-05",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "ARC-AGI-2",
      "score": 1.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-01-20",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 64.88,
      "metric": "accuracy (%)",
      "evidence": "contradicted",
      "evaluator": "Epoch AI",
      "measured_at": "2025-01-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "MATH level 5",
      "score": 93.05,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-01-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 53.29,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-01-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "WeirdML",
      "score": 36.49,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-01-20",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "SimpleBench",
      "score": 17.08,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-01-20",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "METR Time Horizons",
      "score": 51.93,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-01-20",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "Aider polyglot",
      "score": 56.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-01-20",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "Balrog",
      "score": 34.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2025-01-20",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "Fiction.LiveBench",
      "score": 69.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-01-20",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "ARC-AGI",
      "score": 15.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-01-20",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "Lech Mazur Writing",
      "score": 83,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-01-20",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "AIME",
      "score": 79.8,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 4",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "Codeforces rating",
      "score": 2029,
      "metric": "rating",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 4",
      "benchmark_integrity_grade": null
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 71.5,
      "metric": "accuracy (%)",
      "evidence": "contradicted",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 4",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "LiveCodeBench",
      "score": 65.9,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 4",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1",
      "model_slug": "deepseek-r1",
      "developer": "DeepSeek",
      "benchmark": "MATH-500",
      "score": 97.3,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 4",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "ARC AI2",
      "score": 93.73,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-12-26",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "Winogrande",
      "score": 70.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-12-26",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 43.8,
      "metric": "accuracy (%)",
      "evidence": "contradicted",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-26",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "MATH level 5",
      "score": 64.85,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-26",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 15.75,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-26",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "TriviaQA",
      "score": 82.9,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-12-26",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "BBH",
      "score": 83.33,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-12-26",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "SimpleBench",
      "score": 2.68,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2024-12-26",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "METR Time Horizons",
      "score": 47.36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2024-12-26",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 3.02,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-26",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "Aider polyglot",
      "score": 48.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Aider LLM Leaderboards",
      "measured_at": "2024-12-26",
      "source": "Aider LLM Leaderboards",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "HellaSwag",
      "score": 85.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-12-26",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "MMLU",
      "score": 82.93,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-12-26",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "PIQA",
      "score": 69.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-12-26",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "AIME",
      "score": 39.2,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 4",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "Codeforces rating",
      "score": 1134,
      "metric": "rating",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 4",
      "benchmark_integrity_grade": null
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 59.1,
      "metric": "accuracy (%)",
      "evidence": "contradicted",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 4",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "LiveCodeBench",
      "score": 36.2,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 4",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3",
      "model_slug": "deepseek-v3",
      "developer": "DeepSeek",
      "benchmark": "MATH-500",
      "score": 90.2,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 4",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-1.5B",
      "model_slug": "deepseek-ai-deepseek-r1-distill-qwen-1-5b",
      "developer": "DeepSeek",
      "benchmark": "AIME",
      "score": 28.9,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-1.5B",
      "model_slug": "deepseek-ai-deepseek-r1-distill-qwen-1-5b",
      "developer": "DeepSeek",
      "benchmark": "Codeforces rating",
      "score": 954,
      "metric": "rating",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": null
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-1.5B",
      "model_slug": "deepseek-ai-deepseek-r1-distill-qwen-1-5b",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 33.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-1.5B",
      "model_slug": "deepseek-ai-deepseek-r1-distill-qwen-1-5b",
      "developer": "DeepSeek",
      "benchmark": "LiveCodeBench",
      "score": 16.9,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-1.5B",
      "model_slug": "deepseek-ai-deepseek-r1-distill-qwen-1-5b",
      "developer": "DeepSeek",
      "benchmark": "MATH-500",
      "score": 83.9,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3.1-8B",
      "model_slug": "llama-3-1-8b",
      "developer": "Meta AI",
      "benchmark": "GPQA diamond",
      "score": 2.72,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-07-23",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-8B",
      "model_slug": "llama-3-1-8b",
      "developer": "Meta AI",
      "benchmark": "MATH level 5",
      "score": 22.88,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-8B",
      "model_slug": "llama-3-1-8b",
      "developer": "Meta AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 1.57,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-07-23",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-8B",
      "model_slug": "llama-3-1-8b",
      "developer": "Meta AI",
      "benchmark": "DTBench",
      "score": 18.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-07-23",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 3.1-8B",
      "model_slug": "llama-3-1-8b",
      "developer": "Meta AI",
      "benchmark": "WeirdML",
      "score": 1.73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-07-23",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-8B",
      "model_slug": "llama-3-1-8b",
      "developer": "Meta AI",
      "benchmark": "LMCA",
      "score": 6.32,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2024-07-23",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 3.1-8B",
      "model_slug": "llama-3-1-8b",
      "developer": "Meta AI",
      "benchmark": "Balrog",
      "score": 15.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2024-07-23",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-8B",
      "model_slug": "llama-3-1-8b",
      "developer": "Meta AI",
      "benchmark": "GSM8K",
      "score": 82.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-07-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3.1-8B",
      "model_slug": "llama-3-1-8b",
      "developer": "Meta AI",
      "benchmark": "MMLU",
      "score": 41.47,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-07-23",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3.1-8B",
      "model_slug": "llama-3-1-8b",
      "developer": "Meta AI",
      "benchmark": "PIQA",
      "score": 62.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-07-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Mistral 7B v0.1",
      "model_slug": "mistral-7b-v0-1",
      "developer": "Mistral AI",
      "benchmark": "ARC AI2",
      "score": 71.47,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-09-27",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mistral 7B v0.1",
      "model_slug": "mistral-7b-v0-1",
      "developer": "Mistral AI",
      "benchmark": "OpenBookQA",
      "score": 73.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-09-27",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mistral 7B v0.1",
      "model_slug": "mistral-7b-v0-1",
      "developer": "Mistral AI",
      "benchmark": "Winogrande",
      "score": 50.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Training Compute-Optimal Large Language Models",
      "measured_at": "2023-09-27",
      "source": "Training Compute-Optimal Large Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mistral 7B v0.1",
      "model_slug": "mistral-7b-v0-1",
      "developer": "Mistral AI",
      "benchmark": "TriviaQA",
      "score": 75.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-09-27",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Mistral 7B v0.1",
      "model_slug": "mistral-7b-v0-1",
      "developer": "Mistral AI",
      "benchmark": "BBH",
      "score": 41.47,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2023-09-27",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mistral 7B v0.1",
      "model_slug": "mistral-7b-v0-1",
      "developer": "Mistral AI",
      "benchmark": "GSM8K",
      "score": 54.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-09-27",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mistral 7B v0.1",
      "model_slug": "mistral-7b-v0-1",
      "developer": "Mistral AI",
      "benchmark": "HellaSwag",
      "score": 74.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Mixtral of Experts",
      "measured_at": "2023-09-27",
      "source": "Mixtral of Experts",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mistral 7B v0.1",
      "model_slug": "mistral-7b-v0-1",
      "developer": "Mistral AI",
      "benchmark": "ANLI",
      "score": 20.65,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-09-27",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Mistral 7B v0.1",
      "model_slug": "mistral-7b-v0-1",
      "developer": "Mistral AI",
      "benchmark": "MMLU",
      "score": 50,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Mixtral of Experts",
      "measured_at": "2023-09-27",
      "source": "Mixtral of Experts",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mistral 7B v0.1",
      "model_slug": "mistral-7b-v0-1",
      "developer": "Mistral AI",
      "benchmark": "PIQA",
      "score": 66,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Nemotron-4 15B Technical Report",
      "measured_at": "2023-09-27",
      "source": "Nemotron-4 15B Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 2-7B",
      "model_slug": "llama-2-7b",
      "developer": "Meta AI",
      "benchmark": "ARC AI2",
      "score": 27.87,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-07-18",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-7B",
      "model_slug": "llama-2-7b",
      "developer": "Meta AI",
      "benchmark": "OpenBookQA",
      "score": 44.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-07-18",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-7B",
      "model_slug": "llama-2-7b",
      "developer": "Meta AI",
      "benchmark": "Winogrande",
      "score": 38.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-07-18",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-7B",
      "model_slug": "llama-2-7b",
      "developer": "Meta AI",
      "benchmark": "TriviaQA",
      "score": 73.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-07-18",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 2-7B",
      "model_slug": "llama-2-7b",
      "developer": "Meta AI",
      "benchmark": "BBH",
      "score": 18.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Baichuan 2: Open Large-scale Language Models",
      "measured_at": "2023-07-18",
      "source": "Baichuan 2: Open Large-scale Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-7B",
      "model_slug": "llama-2-7b",
      "developer": "Meta AI",
      "benchmark": "GSM8K",
      "score": 16.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2023-07-18",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-7B",
      "model_slug": "llama-2-7b",
      "developer": "Meta AI",
      "benchmark": "HellaSwag",
      "score": 69.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2023-07-18",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-7B",
      "model_slug": "llama-2-7b",
      "developer": "Meta AI",
      "benchmark": "LAMBADA",
      "score": 73.3,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-18",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 2-7B",
      "model_slug": "llama-2-7b",
      "developer": "Meta AI",
      "benchmark": "MMLU",
      "score": 27.73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2023-07-18",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-7B",
      "model_slug": "llama-2-7b",
      "developer": "Meta AI",
      "benchmark": "PIQA",
      "score": 57.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2023-07-18",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Phi-2",
      "model_slug": "phi-2",
      "developer": "Microsoft",
      "benchmark": "ARC AI2",
      "score": 67.87,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-12-12",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Phi-2",
      "model_slug": "phi-2",
      "developer": "Microsoft",
      "benchmark": "OpenBookQA",
      "score": 64.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-12-12",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Phi-2",
      "model_slug": "phi-2",
      "developer": "Microsoft",
      "benchmark": "Winogrande",
      "score": 9.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-12-12",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Phi-2",
      "model_slug": "phi-2",
      "developer": "Microsoft",
      "benchmark": "TriviaQA",
      "score": 45.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-12-12",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Phi-2",
      "model_slug": "phi-2",
      "developer": "Microsoft",
      "benchmark": "BBH",
      "score": 45.87,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-12-12",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Phi-2",
      "model_slug": "phi-2",
      "developer": "Microsoft",
      "benchmark": "HellaSwag",
      "score": 38.13,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-12-12",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Phi-2",
      "model_slug": "phi-2",
      "developer": "Microsoft",
      "benchmark": "ANLI",
      "score": 13.75,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-12-12",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Phi-2",
      "model_slug": "phi-2",
      "developer": "Microsoft",
      "benchmark": "MMLU",
      "score": 44.53,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2023-12-12",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "PowerMoE-3b",
      "model_slug": "ibm-research-powermoe-3b",
      "developer": "IBM Research",
      "benchmark": "MMLU",
      "score": 42.8,
      "metric": "accuracy",
      "evidence": "self_reported",
      "evaluator": "IBM Research",
      "measured_at": "2024-08-14",
      "source": "https://huggingface.co/ibm-research/PowerMoE-3b",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "PowerMoE-3b",
      "model_slug": "ibm-research-powermoe-3b",
      "developer": "IBM Research",
      "benchmark": "GSM8K",
      "score": 25.9,
      "metric": "accuracy",
      "evidence": "self_reported",
      "evaluator": "IBM Research",
      "measured_at": "2024-08-14",
      "source": "https://huggingface.co/ibm-research/PowerMoE-3b",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3.3 70B",
      "model_slug": "llama-3-3-70b",
      "developer": "Meta AI",
      "benchmark": "GPQA diamond",
      "score": 31.17,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-06",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.3 70B",
      "model_slug": "llama-3-3-70b",
      "developer": "Meta AI",
      "benchmark": "MATH level 5",
      "score": 41.6,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-06",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.3 70B",
      "model_slug": "llama-3-3-70b",
      "developer": "Meta AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 5.04,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-06",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.3 70B",
      "model_slug": "llama-3-3-70b",
      "developer": "Meta AI",
      "benchmark": "DTBench",
      "score": 32.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-12-06",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 3.3 70B",
      "model_slug": "llama-3-3-70b",
      "developer": "Meta AI",
      "benchmark": "WeirdML",
      "score": 14.44,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-12-06",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.3 70B",
      "model_slug": "llama-3-3-70b",
      "developer": "Meta AI",
      "benchmark": "LMCA",
      "score": 20.61,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2024-12-06",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 3.3 70B",
      "model_slug": "llama-3-3-70b",
      "developer": "Meta AI",
      "benchmark": "SimpleBench",
      "score": 3.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2024-12-06",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 3.3 70B",
      "model_slug": "llama-3-3-70b",
      "developer": "Meta AI",
      "benchmark": "Balrog",
      "score": 23,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2024-12-06",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.3 70B",
      "model_slug": "llama-3-3-70b",
      "developer": "Meta AI",
      "benchmark": "Fiction.LiveBench",
      "score": 33.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2024-12-06",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.3 70B",
      "model_slug": "llama-3-3-70b",
      "developer": "Meta AI",
      "benchmark": "MMLU",
      "score": 81.73,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-4 Technical Report",
      "measured_at": "2024-12-06",
      "source": "Phi-4 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Phi-4",
      "model_slug": "phi-4",
      "developer": "Microsoft",
      "benchmark": "GPQA diamond",
      "score": 43.14,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Phi-4",
      "model_slug": "phi-4",
      "developer": "Microsoft",
      "benchmark": "MATH level 5",
      "score": 64.94,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Phi-4",
      "model_slug": "phi-4",
      "developer": "Microsoft",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 13.66,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Phi-4",
      "model_slug": "phi-4",
      "developer": "Microsoft",
      "benchmark": "Balrog",
      "score": 11.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2024-12-12",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Phi-4",
      "model_slug": "phi-4",
      "developer": "Microsoft",
      "benchmark": "MMLU",
      "score": 79.73,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-4 Technical Report",
      "measured_at": "2024-12-12",
      "source": "Phi-4 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Phi-4",
      "model_slug": "phi-4",
      "developer": "Microsoft",
      "benchmark": "Lech Mazur Writing",
      "score": 62.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2024-12-12",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "ARC-AGI-2",
      "score": 22.78,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "ProofBench",
      "score": 35,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-06-16",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "GPQA diamond",
      "score": 92.86,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 86.38,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "DTBench",
      "score": 89.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-06-16",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "WeirdML",
      "score": 70.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-06-16",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "LMCA",
      "score": 53.91,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-06-16",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "SimpleQA Verified",
      "score": 34.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-16",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "Chess Puzzles",
      "score": 17.93,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "SimpleBench",
      "score": 50.56,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-06-16",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 62.32,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 29.27,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "EBR-bench",
      "score": 9.52,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "Mystery Game Puzzles",
      "score": 10.77,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "MirrorCode",
      "score": 36.67,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "PostTrainBench",
      "score": 30.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-16",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "Surface Evolver Bench",
      "score": 55.62,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-06-16",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "FrontierCode",
      "score": 24.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-06-16",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "ARC-AGI",
      "score": 77,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-06-16",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "APEX-Agents",
      "score": 45.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-16",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.2",
      "model_slug": "glm-5-2",
      "developer": "Zhipu AI",
      "benchmark": "SWE-Bench verified",
      "score": 78.7,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GLM-5",
      "model_slug": "glm-5",
      "developer": "Zhipu AI",
      "benchmark": "ARC-AGI-2",
      "score": 4.86,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-11",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5",
      "model_slug": "glm-5",
      "developer": "Zhipu AI",
      "benchmark": "GPQA diamond",
      "score": 87.25,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5",
      "model_slug": "glm-5",
      "developer": "Zhipu AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 79.98,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5",
      "model_slug": "glm-5",
      "developer": "Zhipu AI",
      "benchmark": "WeirdML",
      "score": 48.17,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "WeirdML Leaderboard",
      "measured_at": "2026-02-11",
      "source": "WeirdML Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5",
      "model_slug": "glm-5",
      "developer": "Zhipu AI",
      "benchmark": "SimpleBench",
      "score": 43.84,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2026-02-11",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5",
      "model_slug": "glm-5",
      "developer": "Zhipu AI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 28.83,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5",
      "model_slug": "glm-5",
      "developer": "Zhipu AI",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 3.5,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5",
      "model_slug": "glm-5",
      "developer": "Zhipu AI",
      "benchmark": "Terminal Bench",
      "score": 52.42,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2026-02-11",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5",
      "model_slug": "glm-5",
      "developer": "Zhipu AI",
      "benchmark": "CL-bench",
      "score": 18.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-11",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GLM-5",
      "model_slug": "glm-5",
      "developer": "Zhipu AI",
      "benchmark": "ARC-AGI",
      "score": 44.67,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-11",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GLM-5",
      "model_slug": "glm-5",
      "developer": "Zhipu AI",
      "benchmark": "SWE-Bench verified",
      "score": 72.08,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V4-Pro",
      "model_slug": "deepseek-v4-pro",
      "developer": "DeepSeek",
      "benchmark": "ProofBench",
      "score": 16,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-04-24",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V4-Pro",
      "model_slug": "deepseek-v4-pro",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 91.54,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V4-Pro",
      "model_slug": "deepseek-v4-pro",
      "developer": "DeepSeek",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 96.66,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V4-Pro",
      "model_slug": "deepseek-v4-pro",
      "developer": "DeepSeek",
      "benchmark": "DTBench",
      "score": 84.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-04-24",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V4-Pro",
      "model_slug": "deepseek-v4-pro",
      "developer": "DeepSeek",
      "benchmark": "WeirdML",
      "score": 48.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-04-24",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V4-Pro",
      "model_slug": "deepseek-v4-pro",
      "developer": "DeepSeek",
      "benchmark": "LMCA",
      "score": 48.47,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-04-24",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V4-Pro",
      "model_slug": "deepseek-v4-pro",
      "developer": "DeepSeek",
      "benchmark": "SimpleQA Verified",
      "score": 46.99,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-24",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V4-Pro",
      "model_slug": "deepseek-v4-pro",
      "developer": "DeepSeek",
      "benchmark": "Chess Puzzles",
      "score": 14.77,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-24",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V4-Pro",
      "model_slug": "deepseek-v4-pro",
      "developer": "DeepSeek",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 47.65,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V4-Pro",
      "model_slug": "deepseek-v4-pro",
      "developer": "DeepSeek",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 2.44,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V4-Pro",
      "model_slug": "deepseek-v4-pro",
      "developer": "DeepSeek",
      "benchmark": "Mystery Game Puzzles",
      "score": 8.57,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-24",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V4-Pro",
      "model_slug": "deepseek-v4-pro",
      "developer": "DeepSeek",
      "benchmark": "Surface Evolver Bench",
      "score": 40,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-04-24",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V4-Pro",
      "model_slug": "deepseek-v4-pro",
      "developer": "DeepSeek",
      "benchmark": "FrontierCode",
      "score": 17.64,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-04-24",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V4-Pro",
      "model_slug": "deepseek-v4-pro",
      "developer": "DeepSeek",
      "benchmark": "CL-bench Life",
      "score": 13.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-24",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V4-Pro",
      "model_slug": "deepseek-v4-pro",
      "developer": "DeepSeek",
      "benchmark": "SWE-Bench verified",
      "score": 77.64,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-32B",
      "model_slug": "deepseek-ai-deepseek-r1-distill-qwen-32b",
      "developer": "DeepSeek",
      "benchmark": "AIME",
      "score": 72.6,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-32B",
      "model_slug": "deepseek-ai-deepseek-r1-distill-qwen-32b",
      "developer": "DeepSeek",
      "benchmark": "Codeforces rating",
      "score": 1691,
      "metric": "rating",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": null
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-32B",
      "model_slug": "deepseek-ai-deepseek-r1-distill-qwen-32b",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 62.1,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-32B",
      "model_slug": "deepseek-ai-deepseek-r1-distill-qwen-32b",
      "developer": "DeepSeek",
      "benchmark": "LiveCodeBench",
      "score": 57.2,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-32B",
      "model_slug": "deepseek-ai-deepseek-r1-distill-qwen-32b",
      "developer": "DeepSeek",
      "benchmark": "MATH-500",
      "score": 94.3,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-7B",
      "model_slug": "falcon-7b",
      "developer": "Technology Innovation Institute",
      "benchmark": "ARC AI2",
      "score": 30.48,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Falcon2-11B Technical Report",
      "measured_at": "2023-04-24",
      "source": "Falcon2-11B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-7B",
      "model_slug": "falcon-7b",
      "developer": "Technology Innovation Institute",
      "benchmark": "OpenBookQA",
      "score": 35.47,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-04-24",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-7B",
      "model_slug": "falcon-7b",
      "developer": "Technology Innovation Institute",
      "benchmark": "Winogrande",
      "score": 34.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-04-24",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-7B",
      "model_slug": "falcon-7b",
      "developer": "Technology Innovation Institute",
      "benchmark": "TriviaQA",
      "score": 64.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-04-24",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Falcon-7B",
      "model_slug": "falcon-7b",
      "developer": "Technology Innovation Institute",
      "benchmark": "BBH",
      "score": 5.03,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Baichuan 2: Open Large-scale Language Models",
      "measured_at": "2023-04-24",
      "source": "Baichuan 2: Open Large-scale Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-7B",
      "model_slug": "falcon-7b",
      "developer": "Technology Innovation Institute",
      "benchmark": "GSM8K",
      "score": 6.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-04-24",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-7B",
      "model_slug": "falcon-7b",
      "developer": "Technology Innovation Institute",
      "benchmark": "HellaSwag",
      "score": 70.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Falcon2-11B Technical Report",
      "measured_at": "2023-04-24",
      "source": "Falcon2-11B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-7B",
      "model_slug": "falcon-7b",
      "developer": "Technology Innovation Institute",
      "benchmark": "LAMBADA",
      "score": 74.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-04-24",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Falcon-7B",
      "model_slug": "falcon-7b",
      "developer": "Technology Innovation Institute",
      "benchmark": "MMLU",
      "score": 13.33,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Falcon2-11B Technical Report",
      "measured_at": "2023-04-24",
      "source": "Falcon2-11B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-7B",
      "model_slug": "falcon-7b",
      "developer": "Technology Innovation Institute",
      "benchmark": "PIQA",
      "score": 60.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-04-24",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Qwen3-235B-A22B",
      "model_slug": "qwen3-235b-a22b",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 63.48,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-28",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-235B-A22B",
      "model_slug": "qwen3-235b-a22b",
      "developer": "Qwen",
      "benchmark": "MATH level 5",
      "score": 68.86,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-28",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-235B-A22B",
      "model_slug": "qwen3-235b-a22b",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 59.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-04-28",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-235B-A22B",
      "model_slug": "qwen3-235b-a22b",
      "developer": "Qwen",
      "benchmark": "WeirdML",
      "score": 37.28,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "WeirdML Leaderboard",
      "measured_at": "2025-04-28",
      "source": "WeirdML Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-235B-A22B",
      "model_slug": "qwen3-235b-a22b",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 29.41,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-04-28",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-235B-A22B",
      "model_slug": "qwen3-235b-a22b",
      "developer": "Qwen",
      "benchmark": "SimpleBench",
      "score": 17.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-04-28",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-235B-A22B",
      "model_slug": "qwen3-235b-a22b",
      "developer": "Qwen",
      "benchmark": "Aider polyglot",
      "score": 59.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Aider LLM Leaderboards",
      "measured_at": "2025-04-28",
      "source": "Aider LLM Leaderboards",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-235B-A22B",
      "model_slug": "qwen3-235b-a22b",
      "developer": "Qwen",
      "benchmark": "Fiction.LiveBench",
      "score": 67.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-04-28",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-235B-A22B",
      "model_slug": "qwen3-235b-a22b",
      "developer": "Qwen",
      "benchmark": "Lech Mazur Writing",
      "score": 83,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-04-28",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-R1-Distill-Llama-70B",
      "model_slug": "deepseek-ai-deepseek-r1-distill-llama-70b",
      "developer": "DeepSeek",
      "benchmark": "AIME",
      "score": 70,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Llama-70B",
      "model_slug": "deepseek-ai-deepseek-r1-distill-llama-70b",
      "developer": "DeepSeek",
      "benchmark": "Codeforces rating",
      "score": 1633,
      "metric": "rating",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": null
    },
    {
      "model": "DeepSeek-R1-Distill-Llama-70B",
      "model_slug": "deepseek-ai-deepseek-r1-distill-llama-70b",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 65.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Llama-70B",
      "model_slug": "deepseek-ai-deepseek-r1-distill-llama-70b",
      "developer": "DeepSeek",
      "benchmark": "LiveCodeBench",
      "score": 57.5,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Llama-70B",
      "model_slug": "deepseek-ai-deepseek-r1-distill-llama-70b",
      "developer": "DeepSeek",
      "benchmark": "MATH-500",
      "score": 94.5,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "ProofBench",
      "score": 16,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-04-20",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "GPQA diamond",
      "score": 91.37,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 96.11,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "DTBench",
      "score": 84.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-04-20",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "WeirdML",
      "score": 55.86,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-04-20",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "LMCA",
      "score": 43.84,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-04-20",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "SimpleQA Verified",
      "score": 34.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-20",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "Chess Puzzles",
      "score": 15.82,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-20",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 60.2,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 25.64,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "EBR-bench",
      "score": 2.38,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "Mystery Game Puzzles",
      "score": 9.67,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-20",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "Furniture Assembly",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-20",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "ExploitBench",
      "score": 18.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-20",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "OSWorld 2.0",
      "score": 4.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://osworld-v2.xlang.ai/",
      "measured_at": "2026-04-20",
      "source": "https://osworld-v2.xlang.ai/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.6",
      "model_slug": "kimi-k2-6",
      "developer": "NVIDIA",
      "benchmark": "SWE-Bench verified",
      "score": 76.65,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Baichuan 2-7B",
      "model_slug": "baichuan-2-7b",
      "developer": "Baichuan AI",
      "benchmark": "ARC AI2",
      "score": 10,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-20",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Baichuan 2-7B",
      "model_slug": "baichuan-2-7b",
      "developer": "Baichuan AI",
      "benchmark": "BBH",
      "score": 22.13,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-20",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Baichuan 2-7B",
      "model_slug": "baichuan-2-7b",
      "developer": "Baichuan AI",
      "benchmark": "GSM8K",
      "score": 24.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-20",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Baichuan 2-7B",
      "model_slug": "baichuan-2-7b",
      "developer": "Baichuan AI",
      "benchmark": "HellaSwag",
      "score": 57.33,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-20",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Baichuan 2-7B",
      "model_slug": "baichuan-2-7b",
      "developer": "Baichuan AI",
      "benchmark": "LAMBADA",
      "score": 73.3,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-20",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Baichuan 2-7B",
      "model_slug": "baichuan-2-7b",
      "developer": "Baichuan AI",
      "benchmark": "MMLU",
      "score": 38.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Baichuan 2: Open Large-scale Language Models",
      "measured_at": "2023-09-20",
      "source": "Baichuan 2: Open Large-scale Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3.5 Sonnet",
      "model_slug": "claude-3-5-sonnet",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 40.33,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-06-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet",
      "model_slug": "claude-3-5-sonnet",
      "developer": "Anthropic",
      "benchmark": "MATH level 5",
      "score": 51.68,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-06-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet",
      "model_slug": "claude-3-5-sonnet",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 6.43,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-06-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet",
      "model_slug": "claude-3-5-sonnet",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 46.28,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-06-20",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3.5 Sonnet",
      "model_slug": "claude-3-5-sonnet",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 30.97,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-06-20",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet",
      "model_slug": "claude-3-5-sonnet",
      "developer": "Anthropic",
      "benchmark": "Cybench",
      "score": 17.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Cybench leaderboard",
      "measured_at": "2024-06-20",
      "source": "Cybench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet",
      "model_slug": "claude-3-5-sonnet",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 13,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2024-06-20",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3.5 Sonnet",
      "model_slug": "claude-3-5-sonnet",
      "developer": "Anthropic",
      "benchmark": "METR Time Horizons",
      "score": 40.15,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "measured_at": "2024-06-20",
      "source": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3.5 Sonnet",
      "model_slug": "claude-3-5-sonnet",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 1.81,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-06-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3.5 Sonnet",
      "model_slug": "claude-3-5-sonnet",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-06-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet",
      "model_slug": "claude-3-5-sonnet",
      "developer": "Anthropic",
      "benchmark": "VPCT",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2024-06-20",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet",
      "model_slug": "claude-3-5-sonnet",
      "developer": "Anthropic",
      "benchmark": "MMLU",
      "score": 82,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-06-20",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 1.5 Pro (Sept 2024)",
      "model_slug": "gemini-1-5-pro-sept-2024",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI-2",
      "score": 0.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-09-24",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 1.5 Pro (Sept 2024)",
      "model_slug": "gemini-1-5-pro-sept-2024",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 44.76,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Pro (Sept 2024)",
      "model_slug": "gemini-1-5-pro-sept-2024",
      "developer": "Google DeepMind",
      "benchmark": "MATH level 5",
      "score": 70.39,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Pro (Sept 2024)",
      "model_slug": "gemini-1-5-pro-sept-2024",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 22.98,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Pro (Sept 2024)",
      "model_slug": "gemini-1-5-pro-sept-2024",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 31.63,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-09-24",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 1.5 Pro (Sept 2024)",
      "model_slug": "gemini-1-5-pro-sept-2024",
      "developer": "Google DeepMind",
      "benchmark": "CadEval",
      "score": 34,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "CadEval Dashboard",
      "measured_at": "2024-09-24",
      "source": "CadEval Dashboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 1.5 Pro (Sept 2024)",
      "model_slug": "gemini-1-5-pro-sept-2024",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 22.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-09-24",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Pro (Sept 2024)",
      "model_slug": "gemini-1-5-pro-sept-2024",
      "developer": "Google DeepMind",
      "benchmark": "SimpleBench",
      "score": 12.52,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2024-09-24",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 1.5 Pro (Sept 2024)",
      "model_slug": "gemini-1-5-pro-sept-2024",
      "developer": "Google DeepMind",
      "benchmark": "Balrog",
      "score": 21,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2024-09-24",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Pro (Sept 2024)",
      "model_slug": "gemini-1-5-pro-sept-2024",
      "developer": "Google DeepMind",
      "benchmark": "The Agent Company",
      "score": 3.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "TheAgentCompany experiment results github",
      "measured_at": "2024-09-24",
      "source": "TheAgentCompany experiment results github",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Pro (Sept 2024)",
      "model_slug": "gemini-1-5-pro-sept-2024",
      "developer": "Google DeepMind",
      "benchmark": "MMLU",
      "score": 82.53,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-09-24",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 2.0 Flash (Feb 2025)",
      "model_slug": "gemini-2-0-flash-feb-2025",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI-2",
      "score": 1.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-02-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.0 Flash (Feb 2025)",
      "model_slug": "gemini-2-0-flash-feb-2025",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 54.36,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-02-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Flash (Feb 2025)",
      "model_slug": "gemini-2-0-flash-feb-2025",
      "developer": "Google DeepMind",
      "benchmark": "MATH level 5",
      "score": 82.17,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-02-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Flash (Feb 2025)",
      "model_slug": "gemini-2-0-flash-feb-2025",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 31.04,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-02-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Flash (Feb 2025)",
      "model_slug": "gemini-2-0-flash-feb-2025",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 38.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-02-05",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.0 Flash (Feb 2025)",
      "model_slug": "gemini-2-0-flash-feb-2025",
      "developer": "Google DeepMind",
      "benchmark": "CadEval",
      "score": 30,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "CadEval Dashboard",
      "measured_at": "2025-02-05",
      "source": "CadEval Dashboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.0 Flash (Feb 2025)",
      "model_slug": "gemini-2-0-flash-feb-2025",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 25.77,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-02-05",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Flash (Feb 2025)",
      "model_slug": "gemini-2-0-flash-feb-2025",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 3.02,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-02-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.0 Flash (Feb 2025)",
      "model_slug": "gemini-2-0-flash-feb-2025",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 77,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2025-02-05",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Flash (Feb 2025)",
      "model_slug": "gemini-2-0-flash-feb-2025",
      "developer": "Google DeepMind",
      "benchmark": "The Agent Company",
      "score": 11.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "TheAgentCompany experiment results github",
      "measured_at": "2025-02-05",
      "source": "TheAgentCompany experiment results github",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Flash (Feb 2025)",
      "model_slug": "gemini-2-0-flash-feb-2025",
      "developer": "Google DeepMind",
      "benchmark": "Fiction.LiveBench",
      "score": 61.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-02-05",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Flash Thinking (Jan 2025)",
      "model_slug": "gemini-2-0-flash-thinking-jan-2025",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 44.54,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-01-21",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Flash Thinking (Jan 2025)",
      "model_slug": "gemini-2-0-flash-thinking-jan-2025",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 57.74,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-01-21",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Flash Thinking (Jan 2025)",
      "model_slug": "gemini-2-0-flash-thinking-jan-2025",
      "developer": "Google DeepMind",
      "benchmark": "SimpleBench",
      "score": 16.84,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-01-21",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.0 Flash Thinking (Jan 2025)",
      "model_slug": "gemini-2-0-flash-thinking-jan-2025",
      "developer": "Google DeepMind",
      "benchmark": "Aider polyglot",
      "score": 18.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-01-21",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Flash Thinking (Jan 2025)",
      "model_slug": "gemini-2-0-flash-thinking-jan-2025",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 70,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2025-01-21",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Flash Thinking (Jan 2025)",
      "model_slug": "gemini-2-0-flash-thinking-jan-2025",
      "developer": "Google DeepMind",
      "benchmark": "Fiction.LiveBench",
      "score": 52.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-01-21",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Flash Thinking (Jan 2025)",
      "model_slug": "gemini-2-0-flash-thinking-jan-2025",
      "developer": "Google DeepMind",
      "benchmark": "Lech Mazur Writing",
      "score": 73.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-01-21",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 2.0 Pro",
      "model_slug": "gemini-2-0-pro",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 56.47,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-02-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Pro",
      "model_slug": "gemini-2-0-pro",
      "developer": "Google DeepMind",
      "benchmark": "MATH level 5",
      "score": 83.46,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-02-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Pro",
      "model_slug": "gemini-2-0-pro",
      "developer": "Google DeepMind",
      "benchmark": "Aider polyglot",
      "score": 35.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-02-05",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Pro",
      "model_slug": "gemini-2-0-pro",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 77.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2025-02-05",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.0 Pro",
      "model_slug": "gemini-2-0-pro",
      "developer": "Google DeepMind",
      "benchmark": "Fiction.LiveBench",
      "score": 41.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-02-05",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (Apr 2025)",
      "model_slug": "gemini-2-5-flash-apr-2025",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 73.03,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (Apr 2025)",
      "model_slug": "gemini-2-5-flash-apr-2025",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 40.95,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "WeirdML Leaderboard",
      "measured_at": "2025-04-17",
      "source": "WeirdML Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (Apr 2025)",
      "model_slug": "gemini-2-5-flash-apr-2025",
      "developer": "Google DeepMind",
      "benchmark": "Aider polyglot",
      "score": 47.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-04-17",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (Apr 2025)",
      "model_slug": "gemini-2-5-flash-apr-2025",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2025-04-17",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (Apr 2025)",
      "model_slug": "gemini-2-5-flash-apr-2025",
      "developer": "Google DeepMind",
      "benchmark": "VPCT",
      "score": 7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-04-17",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (Apr 2025)",
      "model_slug": "gemini-2-5-flash-apr-2025",
      "developer": "Google DeepMind",
      "benchmark": "Fiction.LiveBench",
      "score": 47.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-04-17",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (Apr 2025)",
      "model_slug": "gemini-2-5-flash-apr-2025",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI",
      "score": 32.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-04-17",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 2.5 Flash (Apr 2025)",
      "model_slug": "gemini-2-5-flash-apr-2025",
      "developer": "Google DeepMind",
      "benchmark": "Lech Mazur Writing",
      "score": 76.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-04-17",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 2.5 Flash (Jun 2025)",
      "model_slug": "gemini-2-5-flash-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 60.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-06-17",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.5 Flash (Jun 2025)",
      "model_slug": "gemini-2-5-flash-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 32.34,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-06-17",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.5 Flash (Jun 2025)",
      "model_slug": "gemini-2-5-flash-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "SimpleBench",
      "score": 29.44,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2025-06-17",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.5 Flash (Jun 2025)",
      "model_slug": "gemini-2-5-flash-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 8.5,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-06-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.5 Flash (Jun 2025)",
      "model_slug": "gemini-2-5-flash-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "Balrog",
      "score": 33.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2025-06-17",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (Jun 2025)",
      "model_slug": "gemini-2-5-flash-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 6.94,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-06-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (Jun 2025)",
      "model_slug": "gemini-2-5-flash-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "Terminal Bench",
      "score": 17.08,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2025-06-17",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (May 2025)",
      "model_slug": "gemini-2-5-flash-may-2025",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI-2",
      "score": 2.54,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-05-20",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.5 Flash (May 2025)",
      "model_slug": "gemini-2-5-flash-may-2025",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 70.8,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (May 2025)",
      "model_slug": "gemini-2-5-flash-may-2025",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 40.95,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "WeirdML Leaderboard",
      "measured_at": "2025-05-20",
      "source": "WeirdML Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (May 2025)",
      "model_slug": "gemini-2-5-flash-may-2025",
      "developer": "Google DeepMind",
      "benchmark": "Aider polyglot",
      "score": 55.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-05-20",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (May 2025)",
      "model_slug": "gemini-2-5-flash-may-2025",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 76,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2025-05-20",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (May 2025)",
      "model_slug": "gemini-2-5-flash-may-2025",
      "developer": "Google DeepMind",
      "benchmark": "Fiction.LiveBench",
      "score": 77.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-05-20",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (May 2025)",
      "model_slug": "gemini-2-5-flash-may-2025",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI",
      "score": 33.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2025-05-20",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 2.5 Flash (Sep 2025)",
      "model_slug": "gemini-2-5-flash-sep-2025",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 41.91,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-09-25",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (Sep 2025)",
      "model_slug": "gemini-2-5-flash-sep-2025",
      "developer": "Google DeepMind",
      "benchmark": "VPCT",
      "score": 19.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-09-25",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (Sep 2025)",
      "model_slug": "gemini-2-5-flash-sep-2025",
      "developer": "Google DeepMind",
      "benchmark": "The Agent Company",
      "score": 41.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "TheAgentCompany leaderboard",
      "measured_at": "2025-09-25",
      "source": "TheAgentCompany leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash (Sep 2025)",
      "model_slug": "gemini-2-5-flash-sep-2025",
      "developer": "Google DeepMind",
      "benchmark": "Terminal Bench",
      "score": 17.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Terminal-Bench v2 Leaderboard",
      "measured_at": "2025-09-25",
      "source": "Terminal-Bench v2 Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash-Lite (Jun 2025)",
      "model_slug": "gemini-2-5-flash-lite-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 38.05,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-06-17",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.5 Flash-Lite (Jun 2025)",
      "model_slug": "gemini-2-5-flash-lite-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 35.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "WeirdML Leaderboard",
      "measured_at": "2025-06-17",
      "source": "WeirdML Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Flash-Lite (Jun 2025)",
      "model_slug": "gemini-2-5-flash-lite-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 21.26,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-06-17",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.5 Flash-Lite (Jun 2025)",
      "model_slug": "gemini-2-5-flash-lite-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "Fiction.LiveBench",
      "score": 47.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-06-17",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI-2",
      "score": 4.86,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-06-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 83.74,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-06-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 84.71,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-06-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 70.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-06-05",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 54.03,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "WeirdML Leaderboard",
      "measured_at": "2025-06-05",
      "source": "WeirdML Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 40.91,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-06-05",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "SimpleBench",
      "score": 54.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2025-06-05",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "METR Time Horizons",
      "score": 55.44,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-06-05",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "Aider polyglot",
      "score": 83.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-06-05",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 81,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2025-06-05",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "VPCT",
      "score": 19.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-06-05",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "GSO-Bench",
      "score": 3.92,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2025-06-05",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "DeepResearch Bench",
      "score": 42.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "DeepResearchBench Leaderboard",
      "measured_at": "2025-06-05",
      "source": "DeepResearchBench Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "Fiction.LiveBench",
      "score": 91.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-06-05",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 25.85,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-06-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-06-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "Remote Labor Index",
      "score": 0.83,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-06-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "Terminal Bench",
      "score": 32.64,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2025-06-05",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "GDPval",
      "score": 23.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-06-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI",
      "score": 41,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-06-05",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "Lech Mazur Writing",
      "score": 83.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-06-05",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 2.5 Pro (Jun 2025)",
      "model_slug": "gemini-2-5-pro-jun-2025",
      "developer": "Google DeepMind",
      "benchmark": "SWE-Bench verified",
      "score": 57.56,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-06-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 42.09,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-10-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "MATH level 5",
      "score": 56.95,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-10-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 8.38,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-10-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 46.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-10-22",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "CadEval",
      "score": 48,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "CadEval Dashboard",
      "measured_at": "2024-10-22",
      "source": "CadEval Dashboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 39.97,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-10-22",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 29.68,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2024-10-22",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "METR Time Horizons",
      "score": 45.24,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "measured_at": "2024-10-22",
      "source": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 3.63,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-10-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "Aider polyglot",
      "score": 51.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2024-10-22",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "Balrog",
      "score": 32.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2024-10-22",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "GeoBench",
      "score": 62,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2024-10-22",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-10-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "VPCT",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2024-10-22",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "GSO-Bench",
      "score": 4.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2024-10-22",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "The Agent Company",
      "score": 24,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "TheAgentCompany experiment results github",
      "measured_at": "2024-10-22",
      "source": "TheAgentCompany experiment results github",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "MMLU",
      "score": 83.07,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-10-22",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3.5 Sonnet (October 2024)",
      "model_slug": "claude-3-5-sonnet-october-2024",
      "developer": "Anthropic",
      "benchmark": "Lech Mazur Writing",
      "score": 80.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2024-10-22",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 2.5 Pro (Mar 2025)",
      "model_slug": "gemini-2-5-pro-mar-2025",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 81.72,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-03-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Mar 2025)",
      "model_slug": "gemini-2-5-pro-mar-2025",
      "developer": "Google DeepMind",
      "benchmark": "MATH level 5",
      "score": 95.56,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-03-31",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Mar 2025)",
      "model_slug": "gemini-2-5-pro-mar-2025",
      "developer": "Google DeepMind",
      "benchmark": "CadEval",
      "score": 64,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "CadEval Dashboard",
      "measured_at": "2025-03-31",
      "source": "CadEval Dashboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.5 Pro (Mar 2025)",
      "model_slug": "gemini-2-5-pro-mar-2025",
      "developer": "Google DeepMind",
      "benchmark": "SimpleBench",
      "score": 41.92,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-03-31",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 2.5 Pro (Mar 2025)",
      "model_slug": "gemini-2-5-pro-mar-2025",
      "developer": "Google DeepMind",
      "benchmark": "Aider polyglot",
      "score": 72.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-03-31",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Mar 2025)",
      "model_slug": "gemini-2-5-pro-mar-2025",
      "developer": "Google DeepMind",
      "benchmark": "Balrog",
      "score": 43.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2025-03-25",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Mar 2025)",
      "model_slug": "gemini-2-5-pro-mar-2025",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 84,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2025-03-25",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Mar 2025)",
      "model_slug": "gemini-2-5-pro-mar-2025",
      "developer": "Google DeepMind",
      "benchmark": "VPCT",
      "score": 22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-03-31",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Mar 2025)",
      "model_slug": "gemini-2-5-pro-mar-2025",
      "developer": "Google DeepMind",
      "benchmark": "Fiction.LiveBench",
      "score": 66.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-03-31",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (Mar 2025)",
      "model_slug": "gemini-2-5-pro-mar-2025",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI",
      "score": 33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-03-31",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 2.5 Pro (Mar 2025)",
      "model_slug": "gemini-2-5-pro-mar-2025",
      "developer": "Google DeepMind",
      "benchmark": "Lech Mazur Writing",
      "score": 80.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-03-25",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 2.5 Pro (May 2025)",
      "model_slug": "gemini-2-5-pro-may-2025",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 57.87,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-06",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (May 2025)",
      "model_slug": "gemini-2-5-pro-may-2025",
      "developer": "Google DeepMind",
      "benchmark": "MATH level 5",
      "score": 95.9,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-06",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (May 2025)",
      "model_slug": "gemini-2-5-pro-may-2025",
      "developer": "Google DeepMind",
      "benchmark": "Aider polyglot",
      "score": 76.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-05-06",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (May 2025)",
      "model_slug": "gemini-2-5-pro-may-2025",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 86,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2025-05-06",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (May 2025)",
      "model_slug": "gemini-2-5-pro-may-2025",
      "developer": "Google DeepMind",
      "benchmark": "VPCT",
      "score": 10.75,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-05-06",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (May 2025)",
      "model_slug": "gemini-2-5-pro-may-2025",
      "developer": "Google DeepMind",
      "benchmark": "The Agent Company",
      "score": 30.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "TheAgentCompany leaderboard",
      "measured_at": "2025-05-06",
      "source": "TheAgentCompany leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (May 2025)",
      "model_slug": "gemini-2-5-pro-may-2025",
      "developer": "Google DeepMind",
      "benchmark": "Fiction.LiveBench",
      "score": 66.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-05-06",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 2.5 Pro (May 2025)",
      "model_slug": "gemini-2-5-pro-may-2025",
      "developer": "Google DeepMind",
      "benchmark": "Lech Mazur Writing",
      "score": 80.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-05-06",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI-2",
      "score": 33.61,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "ProofBench",
      "score": 15,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2025-12-17",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 89.44,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 95.55,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 81.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-12-17",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 61.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-12-17",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 50.74,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-12-17",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "SimpleQA Verified",
      "score": 66.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-17",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "Chess Puzzles",
      "score": 36.87,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "SimpleBench",
      "score": 53.32,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2025-12-17",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "Balrog",
      "score": 48.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2025-12-17",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2025-12-17",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "VPCT",
      "score": 58.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-12-17",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "GSO-Bench",
      "score": 9.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2025-12-17",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "DeepResearch Bench",
      "score": 49.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2025-12-17",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 53.92,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 17.07,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "Mystery Game Puzzles",
      "score": 18.48,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "Terminal Bench",
      "score": 64.27,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2025-12-17",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI",
      "score": 84.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-12-17",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3 Flash",
      "model_slug": "gemini-3-flash",
      "developer": "Google DeepMind",
      "benchmark": "SWE-Bench verified",
      "score": 75.41,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI-2",
      "score": 31.11,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-18",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "ProofBench",
      "score": 20,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2025-11-18",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 93.91,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 91.38,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 69.93,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-11-18",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "SimpleBench",
      "score": 71.68,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2025-11-18",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "METR Time Horizons",
      "score": 70.98,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-18",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 65.96,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "Balrog",
      "score": 58.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2025-11-18",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 84,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2025-11-18",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 31.25,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "VPCT",
      "score": 86.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-11-18",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "GSO-Bench",
      "score": 18.63,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2025-11-18",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "DeepResearch Bench",
      "score": 46.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2025-11-18",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "Remote Labor Index",
      "score": 1.25,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-18",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "Terminal Bench",
      "score": 69.43,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2025-11-18",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "GDPval",
      "score": 40.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-18",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "CL-bench",
      "score": 15.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-18",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI",
      "score": 75,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-11-18",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3 Pro",
      "model_slug": "gemini-3-pro",
      "developer": "Google DeepMind",
      "benchmark": "SWE-Bench verified",
      "score": 72.93,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3.1 Flash-Lite",
      "model_slug": "gemini-3-1-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 78.91,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-03",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Flash-Lite",
      "model_slug": "gemini-3-1-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 79.98,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-03",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Flash-Lite",
      "model_slug": "gemini-3-1-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 61.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-03-03",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.1 Flash-Lite",
      "model_slug": "gemini-3-1-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 52.19,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "WeirdML Leaderboard",
      "measured_at": "2026-03-03",
      "source": "WeirdML Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Flash-Lite",
      "model_slug": "gemini-3-1-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 41.19,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-03-03",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.1 Flash-Lite",
      "model_slug": "gemini-3-1-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 83,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2026-03-03",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Flash-Lite",
      "model_slug": "gemini-3-1-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "DeepResearch Bench",
      "score": 37.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2026-03-03",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3.1 Flash-Lite",
      "model_slug": "gemini-3-1-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 29.18,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-03",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI-2",
      "score": 77.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-19",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "ProofBench",
      "score": 26,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-02-19",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 96.45,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 95.6,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 95.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-02-19",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 72.07,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-02-19",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 63.32,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-02-19",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "SimpleQA Verified",
      "score": 73.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-19",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "Chess Puzzles",
      "score": 48.44,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "SimpleBench",
      "score": 75.52,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-02-19",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "METR Time Horizons",
      "score": 77.03,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "measured_at": "2026-02-19",
      "source": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "Balrog",
      "score": 57,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Balrog Leaderboard",
      "measured_at": "2026-02-19",
      "source": "Balrog Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 84,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2026-02-19",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "GSO-Bench",
      "score": 22.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2026-02-19",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "DeepResearch Bench",
      "score": 47.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2026-02-19",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 62.79,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 26.83,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "EBR-bench",
      "score": 14.29,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "Mystery Game Puzzles",
      "score": 27.3,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "MirrorCode",
      "score": 8.89,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "PostTrainBench",
      "score": 18.11,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-19",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "Furniture Assembly",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-19",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "Terminal Bench",
      "score": 80.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2026-02-19",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "ExploitBench",
      "score": 26.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-19",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "CL-bench",
      "score": 20.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-19",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "CL-bench Life",
      "score": 16.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-19",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI",
      "score": 98,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2026-02-19",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "APEX-Agents",
      "score": 35.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-19",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.1 Pro",
      "model_slug": "gemini-3-1-pro",
      "developer": "Google DeepMind",
      "benchmark": "SWE-Bench verified",
      "score": 75.62,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI-2",
      "score": 72.08,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-05-19",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "ProofBench",
      "score": 31,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-05-19",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 94.17,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 95.55,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 91.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-05-19",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 62.64,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-05-19",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 55.47,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-05-19",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "SimpleQA Verified",
      "score": 66.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-05-19",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "Chess Puzzles",
      "score": 42.13,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-05-19",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "SimpleBench",
      "score": 72.04,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2026-05-19",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 85,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2026-05-19",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 66.11,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 26.83,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "EBR-bench",
      "score": 4.76,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "Mystery Game Puzzles",
      "score": 25.09,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "Surface Evolver Bench",
      "score": 58.13,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-05-19",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI",
      "score": 92.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-05-19",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "APEX-Agents",
      "score": 27.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-05-19",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.5 Flash",
      "model_slug": "gemini-3-5-flash",
      "developer": "Google DeepMind",
      "benchmark": "SWE-Bench verified",
      "score": 79.34,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3.5 Flash-Lite",
      "model_slug": "gemini-3-5-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI-2",
      "score": 10.28,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-21",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash-Lite",
      "model_slug": "gemini-3-5-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "ProofBench",
      "score": 13,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-07-21",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.5 Flash-Lite",
      "model_slug": "gemini-3-5-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 81.02,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-21",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.5 Flash-Lite",
      "model_slug": "gemini-3-5-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 71.08,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-21",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.5 Flash-Lite",
      "model_slug": "gemini-3-5-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 72.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-07-21",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash-Lite",
      "model_slug": "gemini-3-5-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 39,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-07-21",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.5 Flash-Lite",
      "model_slug": "gemini-3-5-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 44.24,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-07-21",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash-Lite",
      "model_slug": "gemini-3-5-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "Chess Puzzles",
      "score": 16.88,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-21",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash-Lite",
      "model_slug": "gemini-3-5-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 27.33,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-21",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash-Lite",
      "model_slug": "gemini-3-5-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-21",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash-Lite",
      "model_slug": "gemini-3-5-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "Mystery Game Puzzles",
      "score": 10.77,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-21",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.5 Flash-Lite",
      "model_slug": "gemini-3-5-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI",
      "score": 53.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-07-21",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3.5 Flash-Lite",
      "model_slug": "gemini-3-5-flash-lite",
      "developer": "Google DeepMind",
      "benchmark": "APEX-Agents",
      "score": 29.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-21",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI-2",
      "score": 60.42,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-21",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "ProofBench",
      "score": 36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-07-21",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 96.01,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-21",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 94.16,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-21",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 92.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-07-21",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 56.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-07-21",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 52.86,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-07-21",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "SimpleQA Verified",
      "score": 66.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-21",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "Chess Puzzles",
      "score": 38.97,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-21",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 62.05,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-21",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 21.95,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-21",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "Mystery Game Puzzles",
      "score": 22.89,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-21",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "MirrorCode",
      "score": 23.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-21",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "FrontierCode",
      "score": 34.37,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-07-21",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "Furniture Assembly",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-21",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI",
      "score": 91.17,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-07-21",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3.6 Flash",
      "model_slug": "gemini-3-6-flash",
      "developer": "Google DeepMind",
      "benchmark": "APEX-Agents",
      "score": 46.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-21",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI-2",
      "score": 84.58,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-13",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "ProofBench",
      "score": 58,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-08-13",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 96.98,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 97.22,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 94.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-08-13",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 59.27,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-08-13",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "SimpleQA Verified",
      "score": 69.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-13",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "Chess Puzzles",
      "score": 64.23,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 75.35,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 36.59,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "Mystery Game Puzzles",
      "score": 30.6,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "MirrorCode",
      "score": 41.11,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-13",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "Remote Labor Index",
      "score": 5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-13",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "FrontierCode",
      "score": 43.59,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-08-13",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "Furniture Assembly",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-13",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "FrontierSWE",
      "score": 20.26,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-08-13",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI",
      "score": 95.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-08-13",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3.7 Flash",
      "model_slug": "gemini-3-7-flash",
      "developer": "Google DeepMind",
      "benchmark": "APEX-Agents",
      "score": 67.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-13",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 0.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-02-24",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 76.02,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-02-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "MATH level 5",
      "score": 91.16,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-02-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 57.74,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-02-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "CadEval",
      "score": 54,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "CadEval Dashboard",
      "measured_at": "2025-02-24",
      "source": "CadEval Dashboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "Cybench",
      "score": 20,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Cybench leaderboard",
      "measured_at": "2025-02-24",
      "source": "Cybench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 35.68,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-02-24",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "METR Time Horizons",
      "score": 59.97,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-02-24",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 7.26,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-02-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "Aider polyglot",
      "score": 64.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-02-24",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "GeoBench",
      "score": 68,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2025-02-24",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "VPCT",
      "score": 8.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-02-24",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "GSO-Bench",
      "score": 3.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GSO Leaderboard",
      "measured_at": "2025-02-24",
      "source": "GSO Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "The Agent Company",
      "score": 30.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "TheAgentCompany leaderboard",
      "measured_at": "2025-02-24",
      "source": "TheAgentCompany leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "DeepResearch Bench",
      "score": 43.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "DeepResearchBench Leaderboard",
      "measured_at": "2025-02-24",
      "source": "DeepResearchBench Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "Fiction.LiveBench",
      "score": 83.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-02-24",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "OSWorld",
      "score": 35.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "OS World Website",
      "measured_at": "2025-02-24",
      "source": "OS World Website",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 28.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-02-24",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "Lech Mazur Writing",
      "score": 81.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-02-24",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3.7 Sonnet",
      "model_slug": "claude-3-7-sonnet",
      "developer": "Anthropic",
      "benchmark": "SWE-Bench verified",
      "score": 60.95,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-02-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI-2",
      "score": 89.17,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "ProofBench",
      "score": 48,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-09-02",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 97.77,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 98.89,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 92.87,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-09-02",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 84.81,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-09-02",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 62.24,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-09-02",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "SimpleQA Verified",
      "score": 69.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "Chess Puzzles",
      "score": 55.81,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "SimpleBench",
      "score": 78.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-09-02",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 86,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2026-09-02",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 72.02,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 21.95,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "Mystery Game Puzzles",
      "score": 41.62,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "MirrorCode",
      "score": 38.89,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "Remote Labor Index",
      "score": 5.83,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "Surface Evolver Bench",
      "score": 76.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-09-02",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "FrontierCode",
      "score": 41.19,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-09-02",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "Furniture Assembly",
      "score": 2.38,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "FrontierSWE",
      "score": 19.63,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-09-02",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML v3",
      "score": 16.19,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-09-02",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI",
      "score": 98.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-09-02",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 3.8 Flash",
      "model_slug": "gemini-3-8-flash",
      "developer": "Google DeepMind",
      "benchmark": "APEX-Agents",
      "score": 64.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI-2",
      "score": 91.25,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-30",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "ProofBench",
      "score": 99,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-09-30",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 99.17,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-30",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-30",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 96,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-09-30",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 74.93,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-09-30",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "SimpleQA Verified",
      "score": 76,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-30",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "Chess Puzzles",
      "score": 56.86,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-30",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "SimpleBench",
      "score": 90.04,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-09-30",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 90.49,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-30",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 78.05,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-30",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "EBR-bench",
      "score": 64.76,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-30",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "Mystery Game Puzzles",
      "score": 60.34,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-30",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "MirrorCode",
      "score": 58.89,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-30",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "PostTrainBench",
      "score": 40.44,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-30",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "Remote Labor Index",
      "score": 17.08,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-30",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "FrontierCode",
      "score": 47.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-09-30",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "Furniture Assembly",
      "score": 60.71,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-30",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "FrontierSWE",
      "score": 54.96,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-09-30",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML v3",
      "score": 45.98,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-09-30",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI",
      "score": 98.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-09-30",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 4 Argon",
      "model_slug": "gemini-4-argon",
      "developer": "Google DeepMind",
      "benchmark": "APEX-Agents",
      "score": 82.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-30",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 4 Carbon",
      "model_slug": "gemini-4-carbon",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI-2",
      "score": 94.17,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Carbon",
      "model_slug": "gemini-4-carbon",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 99.34,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-28",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 4 Carbon",
      "model_slug": "gemini-4-carbon",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-28",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 4 Carbon",
      "model_slug": "gemini-4-carbon",
      "developer": "Google DeepMind",
      "benchmark": "SimpleQA Verified",
      "score": 77.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-28",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 4 Carbon",
      "model_slug": "gemini-4-carbon",
      "developer": "Google DeepMind",
      "benchmark": "Chess Puzzles",
      "score": 76.85,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Carbon",
      "model_slug": "gemini-4-carbon",
      "developer": "Google DeepMind",
      "benchmark": "SimpleBench",
      "score": 91.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-10-28",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Carbon",
      "model_slug": "gemini-4-carbon",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 95.29,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Carbon",
      "model_slug": "gemini-4-carbon",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 95.12,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Carbon",
      "model_slug": "gemini-4-carbon",
      "developer": "Google DeepMind",
      "benchmark": "EBR-bench",
      "score": 87.62,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Carbon",
      "model_slug": "gemini-4-carbon",
      "developer": "Google DeepMind",
      "benchmark": "Mystery Game Puzzles",
      "score": 70.26,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Carbon",
      "model_slug": "gemini-4-carbon",
      "developer": "Google DeepMind",
      "benchmark": "Furniture Assembly",
      "score": 67.86,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 4 Carbon",
      "model_slug": "gemini-4-carbon",
      "developer": "Google DeepMind",
      "benchmark": "FrontierSWE",
      "score": 61.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-10-28",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 4 Carbon",
      "model_slug": "gemini-4-carbon",
      "developer": "Google DeepMind",
      "benchmark": "ARC-AGI",
      "score": 98.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-10-28",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 4 Carbon",
      "model_slug": "gemini-4-carbon",
      "developer": "Google DeepMind",
      "benchmark": "APEX-Agents",
      "score": 84.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-28",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 2 27B",
      "model_slug": "gemma-2-27b",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 15.96,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-06-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 2 27B",
      "model_slug": "gemma-2-27b",
      "developer": "Google DeepMind",
      "benchmark": "MATH level 5",
      "score": 27.89,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-06-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 2 27B",
      "model_slug": "gemma-2-27b",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 1.29,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-06-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 2 27B",
      "model_slug": "gemma-2-27b",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 13.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-06-24",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemma 2 27B",
      "model_slug": "gemma-2-27b",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 8.35,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2024-06-24",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemma 2 27B",
      "model_slug": "gemma-2-27b",
      "developer": "Google DeepMind",
      "benchmark": "MMLU",
      "score": 67.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-06-24",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 2 9B",
      "model_slug": "gemma-2-9b",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 3.42,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-06-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 2 9B",
      "model_slug": "gemma-2-9b",
      "developer": "Google DeepMind",
      "benchmark": "MATH level 5",
      "score": 21.01,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-06-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 2 9B",
      "model_slug": "gemma-2-9b",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 0.46,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-06-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 2 9B",
      "model_slug": "gemma-2-9b",
      "developer": "Google DeepMind",
      "benchmark": "GSM8K",
      "score": 84.9,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-06-24",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 2 9B",
      "model_slug": "gemma-2-9b",
      "developer": "Google DeepMind",
      "benchmark": "MMLU",
      "score": 62.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-06-24",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 2 9B",
      "model_slug": "gemma-2-9b",
      "developer": "Google DeepMind",
      "benchmark": "PIQA",
      "score": 67.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-06-24",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Gemma 2B",
      "model_slug": "gemma-2b",
      "developer": "Google DeepMind",
      "benchmark": "ARC AI2",
      "score": 22.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2024-02-21",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 2B",
      "model_slug": "gemma-2b",
      "developer": "Google DeepMind",
      "benchmark": "Winogrande",
      "score": 30.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2024-02-21",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 2B",
      "model_slug": "gemma-2b",
      "developer": "Google DeepMind",
      "benchmark": "TriviaQA",
      "score": 53.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2024-02-21",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Gemma 2B",
      "model_slug": "gemma-2b",
      "developer": "Google DeepMind",
      "benchmark": "BBH",
      "score": 13.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2024-02-21",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 2B",
      "model_slug": "gemma-2b",
      "developer": "Google DeepMind",
      "benchmark": "GSM8K",
      "score": 17.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2024-02-21",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 2B",
      "model_slug": "gemma-2b",
      "developer": "Google DeepMind",
      "benchmark": "HellaSwag",
      "score": 61.87,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2024-02-21",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 2B",
      "model_slug": "gemma-2b",
      "developer": "Google DeepMind",
      "benchmark": "MMLU",
      "score": 23.07,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2024-02-21",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 2B",
      "model_slug": "gemma-2b",
      "developer": "Google DeepMind",
      "benchmark": "PIQA",
      "score": 54.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2024-02-21",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Gemma 3 27B",
      "model_slug": "gemma-3-27b",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 31.57,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-03-12",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 3 27B",
      "model_slug": "gemma-3-27b",
      "developer": "Google DeepMind",
      "benchmark": "MATH level 5",
      "score": 74.04,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-03-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 3 27B",
      "model_slug": "gemma-3-27b",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 22.42,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-03-12",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 3 27B",
      "model_slug": "gemma-3-27b",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 20.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-03-12",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemma 3 27B",
      "model_slug": "gemma-3-27b",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 14.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-03-12",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemma 3 27B",
      "model_slug": "gemma-3-27b",
      "developer": "Google DeepMind",
      "benchmark": "Aider polyglot",
      "score": 4.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-03-12",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 3 27B",
      "model_slug": "gemma-3-27b",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 52,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2025-03-12",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 3 27B",
      "model_slug": "gemma-3-27b",
      "developer": "Google DeepMind",
      "benchmark": "Fiction.LiveBench",
      "score": 33.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-03-11",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 3 27B",
      "model_slug": "gemma-3-27b",
      "developer": "Google DeepMind",
      "benchmark": "Lech Mazur Writing",
      "score": 79.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-03-12",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 4 26B A4B",
      "model_slug": "gemma-4-26b-a4b",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 66.99,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-02",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 4 26B A4B",
      "model_slug": "gemma-4-26b-a4b",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 82.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-02",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 4 26B A4B",
      "model_slug": "gemma-4-26b-a4b",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 58.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-04-02",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemma 4 26B A4B",
      "model_slug": "gemma-4-26b-a4b",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 35.17,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-04-02",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 4 26B A4B",
      "model_slug": "gemma-4-26b-a4b",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 34.92,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-04-02",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemma 4 31B IT",
      "model_slug": "gemma-4-31b-it",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 70.5,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-02",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 4 31B IT",
      "model_slug": "gemma-4-31b-it",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 73.31,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-02",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 4 31B IT",
      "model_slug": "gemma-4-31b-it",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 71.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-04-02",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemma 4 31B IT",
      "model_slug": "gemma-4-31b-it",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 52.26,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-04-02",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 4 31B IT",
      "model_slug": "gemma-4-31b-it",
      "developer": "Google DeepMind",
      "benchmark": "LMCA",
      "score": 46.18,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-04-02",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemma 4 31B IT",
      "model_slug": "gemma-4-31b-it",
      "developer": "Google DeepMind",
      "benchmark": "SimpleQA Verified",
      "score": 10.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-02",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 4 31B IT",
      "model_slug": "gemma-4-31b-it",
      "developer": "Google DeepMind",
      "benchmark": "Surface Evolver Bench",
      "score": 30.62,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-04-02",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemma 7B",
      "model_slug": "gemma-7b",
      "developer": "Google DeepMind",
      "benchmark": "ARC AI2",
      "score": 71.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-02-21",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 7B",
      "model_slug": "gemma-7b",
      "developer": "Google DeepMind",
      "benchmark": "OpenBookQA",
      "score": 71.47,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-02-21",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 7B",
      "model_slug": "gemma-7b",
      "developer": "Google DeepMind",
      "benchmark": "Winogrande",
      "score": 58,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2024-02-21",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 7B",
      "model_slug": "gemma-7b",
      "developer": "Google DeepMind",
      "benchmark": "TriviaQA",
      "score": 72.3,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-02-21",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Gemma 7B",
      "model_slug": "gemma-7b",
      "developer": "Google DeepMind",
      "benchmark": "BBH",
      "score": 40.13,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2024-02-21",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 7B",
      "model_slug": "gemma-7b",
      "developer": "Google DeepMind",
      "benchmark": "GSM8K",
      "score": 46.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2024-02-21",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 7B",
      "model_slug": "gemma-7b",
      "developer": "Google DeepMind",
      "benchmark": "HellaSwag",
      "score": 76.27,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2024-02-21",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 7B",
      "model_slug": "gemma-7b",
      "developer": "Google DeepMind",
      "benchmark": "ANLI",
      "score": 23.05,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-02-21",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Gemma 7B",
      "model_slug": "gemma-7b",
      "developer": "Google DeepMind",
      "benchmark": "MMLU",
      "score": 54.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-02-21",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemma 7B",
      "model_slug": "gemma-7b",
      "developer": "Google DeepMind",
      "benchmark": "PIQA",
      "score": 62.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2024-02-21",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 89.17,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "ProofBench",
      "score": 95,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-06-09",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 84.53,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 97.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-06-09",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 91.94,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-06-09",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 71.94,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-06-09",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 70.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 44.23,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 78.28,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-06-09",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "GeoBench",
      "score": 89,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2026-06-09",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "GSO-Bench",
      "score": 78.43,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2026-06-09",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 91.6,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 90.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "EBR-bench",
      "score": 39.52,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "Mystery Game Puzzles",
      "score": 47.12,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "MirrorCode",
      "score": 70,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "PostTrainBench",
      "score": 37.27,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-09",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "Remote Labor Index",
      "score": 16.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "Surface Evolver Bench",
      "score": 95,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-06-09",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "FrontierCode",
      "score": 53.48,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-06-09",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "Furniture Assembly",
      "score": 8.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "FrontierSWE",
      "score": 46.96,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-06-09",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "WeirdML v3",
      "score": 30.96,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-06-09",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 98.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-06-09",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Fable 5",
      "model_slug": "claude-fable-5",
      "developer": "Anthropic",
      "benchmark": "APEX-Agents",
      "score": 63.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 3",
      "model_slug": "grok-3",
      "developer": "xAI",
      "benchmark": "ARC-AGI-2",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 3",
      "model_slug": "grok-3",
      "developer": "xAI",
      "benchmark": "GPQA diamond",
      "score": 70.5,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 3",
      "model_slug": "grok-3",
      "developer": "xAI",
      "benchmark": "MATH level 5",
      "score": 88.75,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 3",
      "model_slug": "grok-3",
      "developer": "xAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 55.51,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 3",
      "model_slug": "grok-3",
      "developer": "xAI",
      "benchmark": "WeirdML",
      "score": 37.24,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-04-09",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 3",
      "model_slug": "grok-3",
      "developer": "xAI",
      "benchmark": "SimpleBench",
      "score": 23.32,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-04-09",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 3",
      "model_slug": "grok-3",
      "developer": "xAI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 6.65,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 3",
      "model_slug": "grok-3",
      "developer": "xAI",
      "benchmark": "Aider polyglot",
      "score": 53.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-04-09",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 3",
      "model_slug": "grok-3",
      "developer": "xAI",
      "benchmark": "Balrog",
      "score": 29.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2025-04-09",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 3",
      "model_slug": "grok-3",
      "developer": "xAI",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 3",
      "model_slug": "grok-3",
      "developer": "xAI",
      "benchmark": "Fiction.LiveBench",
      "score": 58.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-04-09",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 3",
      "model_slug": "grok-3",
      "developer": "xAI",
      "benchmark": "ARC-AGI",
      "score": 5.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-04-09",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Grok 3",
      "model_slug": "grok-3",
      "developer": "xAI",
      "benchmark": "Lech Mazur Writing",
      "score": 76.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-04-09",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "ARC-AGI-2",
      "score": 15.97,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "GPQA diamond",
      "score": 86.11,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 83.98,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "WeirdML",
      "score": 45.73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-07-09",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "Cybench",
      "score": 43,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Grok 4 Model Card",
      "measured_at": "2025-07-09",
      "source": "Grok 4 Model Card",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "SimpleBench",
      "score": 52.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2025-07-09",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "METR Time Horizons",
      "score": 66.58,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-07-09",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 34.48,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "Aider polyglot",
      "score": 79.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Aider LLM Leaderboards",
      "measured_at": "2025-07-09",
      "source": "Aider LLM Leaderboards",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "Balrog",
      "score": 43.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2025-07-09",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "GeoBench",
      "score": 45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2025-07-09",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 3.47,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "DeepResearch Bench",
      "score": 47.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2025-07-09",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "Fiction.LiveBench",
      "score": 94.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-07-09",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "Terminal Bench",
      "score": 27.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Terminal-Bench v2 Leaderboard",
      "measured_at": "2025-07-09",
      "source": "Terminal-Bench v2 Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "GDPval",
      "score": 21.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "ARC-AGI",
      "score": 66.67,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Grok 4",
      "model_slug": "grok-4",
      "developer": "xAI",
      "benchmark": "Lech Mazur Writing",
      "score": 76.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-07-09",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Grok 4 Fast",
      "model_slug": "grok-4-fast",
      "developer": "xAI",
      "benchmark": "ARC-AGI-2",
      "score": 5.28,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-09-19",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4 Fast",
      "model_slug": "grok-4-fast",
      "developer": "xAI",
      "benchmark": "DTBench",
      "score": 71.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-09-19",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4 Fast",
      "model_slug": "grok-4-fast",
      "developer": "xAI",
      "benchmark": "WeirdML",
      "score": 42.86,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-09-19",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4 Fast",
      "model_slug": "grok-4-fast",
      "developer": "xAI",
      "benchmark": "Cybench",
      "score": 30,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Grok 4 Fast Model Card",
      "measured_at": "2025-09-19",
      "source": "Grok 4 Fast Model Card",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4 Fast",
      "model_slug": "grok-4-fast",
      "developer": "xAI",
      "benchmark": "Fiction.LiveBench",
      "score": 94.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-09-19",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4 Fast",
      "model_slug": "grok-4-fast",
      "developer": "xAI",
      "benchmark": "ARC-AGI",
      "score": 48.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-09-19",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Grok 4.20",
      "model_slug": "grok-4-20",
      "developer": "xAI",
      "benchmark": "ARC-AGI-2",
      "score": 65.14,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.20",
      "model_slug": "grok-4-20",
      "developer": "xAI",
      "benchmark": "ProofBench",
      "score": 14,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-02-17",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.20",
      "model_slug": "grok-4-20",
      "developer": "xAI",
      "benchmark": "GPQA diamond",
      "score": 89.35,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.20",
      "model_slug": "grok-4-20",
      "developer": "xAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 92.21,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.20",
      "model_slug": "grok-4-20",
      "developer": "xAI",
      "benchmark": "DTBench",
      "score": 83.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-02-17",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.20",
      "model_slug": "grok-4-20",
      "developer": "xAI",
      "benchmark": "WeirdML",
      "score": 52.26,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-02-17",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.20",
      "model_slug": "grok-4-20",
      "developer": "xAI",
      "benchmark": "LMCA",
      "score": 45.48,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-02-17",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.20",
      "model_slug": "grok-4-20",
      "developer": "xAI",
      "benchmark": "SimpleQA Verified",
      "score": 30.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-17",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.20",
      "model_slug": "grok-4-20",
      "developer": "xAI",
      "benchmark": "Chess Puzzles",
      "score": 25.29,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.20",
      "model_slug": "grok-4-20",
      "developer": "xAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 47.28,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.20",
      "model_slug": "grok-4-20",
      "developer": "xAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 17.07,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.20",
      "model_slug": "grok-4-20",
      "developer": "xAI",
      "benchmark": "Terminal Bench",
      "score": 57.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2026-02-17",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.20",
      "model_slug": "grok-4-20",
      "developer": "xAI",
      "benchmark": "CL-bench",
      "score": 22.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-17",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Grok 4.20",
      "model_slug": "grok-4-20",
      "developer": "xAI",
      "benchmark": "CL-bench Life",
      "score": 11.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-17",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.20",
      "model_slug": "grok-4-20",
      "developer": "xAI",
      "benchmark": "ARC-AGI",
      "score": 89.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-02-17",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Grok 4.3 Beta",
      "model_slug": "grok-4-3-beta",
      "developer": "xAI",
      "benchmark": "ProofBench",
      "score": 11,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-04-17",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.3 Beta",
      "model_slug": "grok-4-3-beta",
      "developer": "xAI",
      "benchmark": "GPQA diamond",
      "score": 88.65,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.3 Beta",
      "model_slug": "grok-4-3-beta",
      "developer": "xAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 93.33,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.3 Beta",
      "model_slug": "grok-4-3-beta",
      "developer": "xAI",
      "benchmark": "DTBench",
      "score": 84.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-04-17",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.3 Beta",
      "model_slug": "grok-4-3-beta",
      "developer": "xAI",
      "benchmark": "WeirdML",
      "score": 49.89,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-04-17",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.3 Beta",
      "model_slug": "grok-4-3-beta",
      "developer": "xAI",
      "benchmark": "LMCA",
      "score": 45.09,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-04-17",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.3 Beta",
      "model_slug": "grok-4-3-beta",
      "developer": "xAI",
      "benchmark": "SimpleQA Verified",
      "score": 33.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-17",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.3 Beta",
      "model_slug": "grok-4-3-beta",
      "developer": "xAI",
      "benchmark": "Chess Puzzles",
      "score": 13.72,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.3 Beta",
      "model_slug": "grok-4-3-beta",
      "developer": "xAI",
      "benchmark": "GeoBench",
      "score": 56,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2026-04-17",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.3 Beta",
      "model_slug": "grok-4-3-beta",
      "developer": "xAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 45.06,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.3 Beta",
      "model_slug": "grok-4-3-beta",
      "developer": "xAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 14.63,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "ARC-AGI-2",
      "score": 52.64,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-08",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "ProofBench",
      "score": 31,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-07-08",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "GPQA diamond",
      "score": 95.05,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-08",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 97.78,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-08",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "DTBench",
      "score": 94.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-07-08",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "WeirdML",
      "score": 46.43,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-07-08",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "LMCA",
      "score": 53.18,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-07-08",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "SimpleQA Verified",
      "score": 48.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-08",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "Chess Puzzles",
      "score": 31.61,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-08",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "SimpleBench",
      "score": 64,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-07-08",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 60.2,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-08",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 24.39,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-08",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "MirrorCode",
      "score": 32.22,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-08",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "PostTrainBench",
      "score": 22.62,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-08",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "Surface Evolver Bench",
      "score": 74.38,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-07-08",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "FrontierCode",
      "score": 42.44,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-07-08",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "Furniture Assembly",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-08",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "ARC-AGI",
      "score": 87.17,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-07-08",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Grok 4.5",
      "model_slug": "grok-4-5",
      "developer": "xAI",
      "benchmark": "APEX-Agents",
      "score": 56.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-08",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "ARC-AGI-2",
      "score": 67.08,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-12",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "ProofBench",
      "score": 51,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-08-12",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "GPQA diamond",
      "score": 95.84,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 99.17,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "DTBench",
      "score": 95.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-08-12",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "WeirdML",
      "score": 67.29,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-08-12",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "LMCA",
      "score": 57,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-08-12",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "SimpleQA Verified",
      "score": 49.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-12",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "Chess Puzzles",
      "score": 29.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-12",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "SimpleBench",
      "score": 71.08,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-08-12",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 69.44,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 31.71,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "EBR-bench",
      "score": 30.48,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 27.3,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "MirrorCode",
      "score": 41.11,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-12",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "FrontierCode",
      "score": 48.01,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-08-12",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "Furniture Assembly",
      "score": 14.29,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-12",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "FrontierSWE",
      "score": 25.29,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-08-12",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "ARC-AGI",
      "score": 87.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-08-12",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Grok 4.6",
      "model_slug": "grok-4-6",
      "developer": "xAI",
      "benchmark": "APEX-Agents",
      "score": 65.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-12",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "ARC-AGI-2",
      "score": 61.39,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-21",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "ProofBench",
      "score": 26,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-09-21",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "GPQA diamond",
      "score": 94.08,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-21",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 98.05,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-21",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "DTBench",
      "score": 93.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-09-21",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "LMCA",
      "score": 58.09,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-09-21",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "SimpleQA Verified",
      "score": 56,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-21",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "Chess Puzzles",
      "score": 37.92,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-21",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "GeoBench",
      "score": 61,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2026-09-21",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 55.77,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-21",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 17.07,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-21",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 21.79,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-21",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "MirrorCode",
      "score": 54.44,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-21",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "FrontierCode",
      "score": 47.59,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-09-21",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "Furniture Assembly",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-21",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "FrontierSWE",
      "score": 29.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-09-21",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "WeirdML v3",
      "score": 20.81,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-09-21",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "ARC-AGI",
      "score": 90.17,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-09-21",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Grok 4.7",
      "model_slug": "grok-4-7",
      "developer": "xAI",
      "benchmark": "APEX-Agents",
      "score": 54.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-21",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok-2 (Dec 2024)",
      "model_slug": "grok-2-dec-2024",
      "developer": "xAI",
      "benchmark": "GPQA diamond",
      "score": 39.98,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok-2 (Dec 2024)",
      "model_slug": "grok-2-dec-2024",
      "developer": "xAI",
      "benchmark": "MATH level 5",
      "score": 63.52,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok-2 (Dec 2024)",
      "model_slug": "grok-2-dec-2024",
      "developer": "xAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 11.44,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok-2 (Dec 2024)",
      "model_slug": "grok-2-dec-2024",
      "developer": "xAI",
      "benchmark": "DTBench",
      "score": 42.02,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-12-12",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok-2 (Dec 2024)",
      "model_slug": "grok-2-dec-2024",
      "developer": "xAI",
      "benchmark": "WeirdML",
      "score": 22.24,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-12-12",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok-2 (Dec 2024)",
      "model_slug": "grok-2-dec-2024",
      "developer": "xAI",
      "benchmark": "SimpleBench",
      "score": 7.24,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2024-12-12",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok-2 (Dec 2024)",
      "model_slug": "grok-2-dec-2024",
      "developer": "xAI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 1.21,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok-2 (Dec 2024)",
      "model_slug": "grok-2-dec-2024",
      "developer": "xAI",
      "benchmark": "Lech Mazur Writing",
      "score": 63.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2024-12-12",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Grok-3 mini",
      "model_slug": "grok-3-mini",
      "developer": "xAI",
      "benchmark": "ARC-AGI-2",
      "score": 0.42,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-06-24",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok-3 mini",
      "model_slug": "grok-3-mini",
      "developer": "xAI",
      "benchmark": "GPQA diamond",
      "score": 71.2,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok-3 mini",
      "model_slug": "grok-3-mini",
      "developer": "xAI",
      "benchmark": "MATH level 5",
      "score": 90.94,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok-3 mini",
      "model_slug": "grok-3-mini",
      "developer": "xAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 77.76,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok-3 mini",
      "model_slug": "grok-3-mini",
      "developer": "xAI",
      "benchmark": "WeirdML",
      "score": 42.58,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "WeirdML Leaderboard",
      "measured_at": "2025-06-24",
      "source": "WeirdML Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok-3 mini",
      "model_slug": "grok-3-mini",
      "developer": "xAI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 10.28,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Grok-3 mini",
      "model_slug": "grok-3-mini",
      "developer": "xAI",
      "benchmark": "Aider polyglot",
      "score": 49.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-04-09",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok-3 mini",
      "model_slug": "grok-3-mini",
      "developer": "xAI",
      "benchmark": "Fiction.LiveBench",
      "score": 66.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-04-09",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Grok-3 mini",
      "model_slug": "grok-3-mini",
      "developer": "xAI",
      "benchmark": "ARC-AGI",
      "score": 16.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-06-24",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Grok-3 mini",
      "model_slug": "grok-3-mini",
      "developer": "xAI",
      "benchmark": "Lech Mazur Writing",
      "score": 73.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-04-09",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 90,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "ProofBench",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-09-01",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 96,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-09-01",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 92.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-09-01",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 77.01,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-09-01",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 70.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 45.29,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 83.92,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-09-01",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "GeoBench",
      "score": 84,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2026-09-01",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "GSO-Bench",
      "score": 88.24,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2026-09-01",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 94.92,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 87.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "EBR-bench",
      "score": 57.14,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "Mystery Game Puzzles",
      "score": 53.73,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "MirrorCode",
      "score": 93.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "PostTrainBench",
      "score": 44.65,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "Remote Labor Index",
      "score": 17.92,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "FrontierCode",
      "score": 50.91,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-09-01",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "Furniture Assembly",
      "score": 57.14,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "FrontierSWE",
      "score": 56.29,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-09-01",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "WeirdML v3",
      "score": 48.36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-09-01",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "SpireClimb",
      "score": 2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 97.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-09-01",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Fable 5.1",
      "model_slug": "claude-fable-5-1",
      "developer": "Anthropic",
      "benchmark": "APEX-Agents",
      "score": 68.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "INTELLECT-1",
      "model_slug": "intellect-1",
      "developer": "Prime Intellect",
      "benchmark": "ARC AI2",
      "score": 39.36,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "INTELLECT-1 Technical Report",
      "measured_at": "2024-11-29",
      "source": "INTELLECT-1 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "INTELLECT-1",
      "model_slug": "intellect-1",
      "developer": "Prime Intellect",
      "benchmark": "Winogrande",
      "score": 31.64,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "INTELLECT-1 Technical Report",
      "measured_at": "2024-11-29",
      "source": "INTELLECT-1 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "INTELLECT-1",
      "model_slug": "intellect-1",
      "developer": "Prime Intellect",
      "benchmark": "BBH",
      "score": 13.13,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "INTELLECT-1 Technical Report",
      "measured_at": "2024-11-29",
      "source": "INTELLECT-1 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "INTELLECT-1",
      "model_slug": "intellect-1",
      "developer": "Prime Intellect",
      "benchmark": "GSM8K",
      "score": 38.58,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "INTELLECT-1 Technical Report",
      "measured_at": "2024-11-29",
      "source": "INTELLECT-1 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "INTELLECT-1",
      "model_slug": "intellect-1",
      "developer": "Prime Intellect",
      "benchmark": "HellaSwag",
      "score": 61.89,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "INTELLECT-1 Technical Report",
      "measured_at": "2024-11-29",
      "source": "INTELLECT-1 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "INTELLECT-1",
      "model_slug": "intellect-1",
      "developer": "Prime Intellect",
      "benchmark": "MMLU",
      "score": 33.19,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "INTELLECT-1 Technical Report",
      "measured_at": "2024-11-29",
      "source": "INTELLECT-1 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "ARC-AGI-2",
      "score": 36.53,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-15",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "ProofBench",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-07-15",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "GPQA diamond",
      "score": 87.86,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-15",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 88.88,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-15",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "DTBench",
      "score": 79.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-07-15",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "WeirdML",
      "score": 32.29,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-07-15",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "LMCA",
      "score": 44.21,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-07-15",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "SimpleQA Verified",
      "score": 40.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-15",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "Chess Puzzles",
      "score": 16.88,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-15",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "SimpleBench",
      "score": 40,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-07-15",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 35.09,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-15",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 4.88,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-15",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "FrontierCode",
      "score": 14,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-07-15",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "FrontierSWE",
      "score": 4.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-07-15",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "WeirdML v3",
      "score": 4.87,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-07-15",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "ARC-AGI",
      "score": 79.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-07-15",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Inkling",
      "model_slug": "inkling",
      "developer": null,
      "benchmark": "APEX-Agents",
      "score": 33.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-15",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Inkling-Small",
      "model_slug": "inkling-small",
      "developer": null,
      "benchmark": "ARC-AGI-2",
      "score": 40.14,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-15",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Inkling-Small",
      "model_slug": "inkling-small",
      "developer": null,
      "benchmark": "ProofBench",
      "score": 6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-07-15",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Inkling-Small",
      "model_slug": "inkling-small",
      "developer": null,
      "benchmark": "GPQA diamond",
      "score": 88.21,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-15",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Inkling-Small",
      "model_slug": "inkling-small",
      "developer": null,
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 89.99,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-15",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Inkling-Small",
      "model_slug": "inkling-small",
      "developer": null,
      "benchmark": "SimpleQA Verified",
      "score": 19.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-15",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Inkling-Small",
      "model_slug": "inkling-small",
      "developer": null,
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 48.75,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-15",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Inkling-Small",
      "model_slug": "inkling-small",
      "developer": null,
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 17.07,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-15",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Inkling-Small",
      "model_slug": "inkling-small",
      "developer": null,
      "benchmark": "Mystery Game Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-15",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Inkling-Small",
      "model_slug": "inkling-small",
      "developer": null,
      "benchmark": "ARC-AGI",
      "score": 84,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-07-15",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Kimi K2 (Jul 2025)",
      "model_slug": "kimi-k2-jul-2025",
      "developer": "Moonshot AI",
      "benchmark": "WeirdML",
      "score": 39.36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "WeirdML Leaderboard",
      "measured_at": "2025-07-12",
      "source": "WeirdML Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2 (Jul 2025)",
      "model_slug": "kimi-k2-jul-2025",
      "developer": "Moonshot AI",
      "benchmark": "SimpleBench",
      "score": 11.56,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2025-07-12",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2 (Jul 2025)",
      "model_slug": "kimi-k2-jul-2025",
      "developer": "Moonshot AI",
      "benchmark": "Aider polyglot",
      "score": 59.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Aider LLM Leaderboards",
      "measured_at": "2025-07-12",
      "source": "Aider LLM Leaderboards",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2 (Jul 2025)",
      "model_slug": "kimi-k2-jul-2025",
      "developer": "Moonshot AI",
      "benchmark": "GSO-Bench",
      "score": 4.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2025-07-12",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2 (Jul 2025)",
      "model_slug": "kimi-k2-jul-2025",
      "developer": "Moonshot AI",
      "benchmark": "Fiction.LiveBench",
      "score": 61.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-07-12",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2 (Jul 2025)",
      "model_slug": "kimi-k2-jul-2025",
      "developer": "Moonshot AI",
      "benchmark": "Terminal Bench",
      "score": 27.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Terminal-Bench v2 Leaderboard",
      "measured_at": "2025-07-12",
      "source": "Terminal-Bench v2 Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2 (Jul 2025)",
      "model_slug": "kimi-k2-jul-2025",
      "developer": "Moonshot AI",
      "benchmark": "Lech Mazur Writing",
      "score": 85.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-07-12",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Kimi K2 Thinking",
      "model_slug": "kimi-k2-thinking",
      "developer": "Moonshot AI",
      "benchmark": "GPQA diamond",
      "score": 82.25,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-06",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2 Thinking",
      "model_slug": "kimi-k2-thinking",
      "developer": "Moonshot AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 83.04,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-06",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2 Thinking",
      "model_slug": "kimi-k2-thinking",
      "developer": "Moonshot AI",
      "benchmark": "WeirdML",
      "score": 42.79,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-11-06",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2 Thinking",
      "model_slug": "kimi-k2-thinking",
      "developer": "Moonshot AI",
      "benchmark": "SimpleBench",
      "score": 27.52,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2025-11-06",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2 Thinking",
      "model_slug": "kimi-k2-thinking",
      "developer": "Moonshot AI",
      "benchmark": "METR Time Horizons",
      "score": 59.19,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-11-06",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Kimi K2 Thinking",
      "model_slug": "kimi-k2-thinking",
      "developer": "Moonshot AI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 37.55,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-06",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2 Thinking",
      "model_slug": "kimi-k2-thinking",
      "developer": "Moonshot AI",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-06",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2 Thinking",
      "model_slug": "kimi-k2-thinking",
      "developer": "Moonshot AI",
      "benchmark": "Terminal Bench",
      "score": 35.73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2025-11-06",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2 Thinking",
      "model_slug": "kimi-k2-thinking",
      "developer": "Moonshot AI",
      "benchmark": "CL-bench",
      "score": 17.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-06",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Kimi K2.5",
      "model_slug": "kimi-k2-5",
      "developer": "Moonshot AI",
      "benchmark": "ARC-AGI-2",
      "score": 11.81,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-01-27",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2.5",
      "model_slug": "kimi-k2-5",
      "developer": "Moonshot AI",
      "benchmark": "GPQA diamond",
      "score": 86.94,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-01-27",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.5",
      "model_slug": "kimi-k2-5",
      "developer": "Moonshot AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 92.19,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-01-27",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.5",
      "model_slug": "kimi-k2-5",
      "developer": "Moonshot AI",
      "benchmark": "WeirdML",
      "score": 45.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-01-27",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.5",
      "model_slug": "kimi-k2-5",
      "developer": "Moonshot AI",
      "benchmark": "SimpleQA Verified",
      "score": 34.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-01-27",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.5",
      "model_slug": "kimi-k2-5",
      "developer": "Moonshot AI",
      "benchmark": "SimpleBench",
      "score": 36.16,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2026-01-27",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2.5",
      "model_slug": "kimi-k2-5",
      "developer": "Moonshot AI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 48.95,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-01-27",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2.5",
      "model_slug": "kimi-k2-5",
      "developer": "Moonshot AI",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 7,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-01-27",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.5",
      "model_slug": "kimi-k2-5",
      "developer": "Moonshot AI",
      "benchmark": "Fiction.LiveBench",
      "score": 86.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2026-01-27",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.5",
      "model_slug": "kimi-k2-5",
      "developer": "Moonshot AI",
      "benchmark": "OSWorld",
      "score": 63.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "OS World Website",
      "measured_at": "2026-01-27",
      "source": "OS World Website",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Kimi K2.5",
      "model_slug": "kimi-k2-5",
      "developer": "Moonshot AI",
      "benchmark": "Terminal Bench",
      "score": 43.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2026-01-27",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.5",
      "model_slug": "kimi-k2-5",
      "developer": "Moonshot AI",
      "benchmark": "CL-bench",
      "score": 19.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-01-27",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Kimi K2.5",
      "model_slug": "kimi-k2-5",
      "developer": "Moonshot AI",
      "benchmark": "CL-bench Life",
      "score": 13.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-01-27",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.5",
      "model_slug": "kimi-k2-5",
      "developer": "Moonshot AI",
      "benchmark": "ARC-AGI",
      "score": 65.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-01-27",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Kimi K2.5",
      "model_slug": "kimi-k2-5",
      "developer": "Moonshot AI",
      "benchmark": "SWE-Bench verified",
      "score": 73.76,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-01-27",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Kimi K2.7 Code",
      "model_slug": "kimi-k2-7-code",
      "developer": "Moonshot AI",
      "benchmark": "GPQA diamond",
      "score": 87.33,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.7 Code",
      "model_slug": "kimi-k2-7-code",
      "developer": "Moonshot AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 95.55,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.7 Code",
      "model_slug": "kimi-k2-7-code",
      "developer": "Moonshot AI",
      "benchmark": "WeirdML",
      "score": 54.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-06-12",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.7 Code",
      "model_slug": "kimi-k2-7-code",
      "developer": "Moonshot AI",
      "benchmark": "SimpleQA Verified",
      "score": 36.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-12",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.7 Code",
      "model_slug": "kimi-k2-7-code",
      "developer": "Moonshot AI",
      "benchmark": "Chess Puzzles",
      "score": 9.51,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-12",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2.7 Code",
      "model_slug": "kimi-k2-7-code",
      "developer": "Moonshot AI",
      "benchmark": "SimpleBench",
      "score": 49.48,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-06-12",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2.7 Code",
      "model_slug": "kimi-k2-7-code",
      "developer": "Moonshot AI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 56.88,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2.7 Code",
      "model_slug": "kimi-k2-7-code",
      "developer": "Moonshot AI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 12.2,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K2.7 Code",
      "model_slug": "kimi-k2-7-code",
      "developer": "Moonshot AI",
      "benchmark": "Surface Evolver Bench",
      "score": 48.75,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-06-12",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.7 Code",
      "model_slug": "kimi-k2-7-code",
      "developer": "Moonshot AI",
      "benchmark": "FrontierCode",
      "score": 30.06,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-06-12",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K2.7 Code",
      "model_slug": "kimi-k2-7-code",
      "developer": "Moonshot AI",
      "benchmark": "APEX-Agents",
      "score": 37.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-12",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "ARC-AGI-2",
      "score": 60.42,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "ProofBench",
      "score": 87,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-07-16",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "GPQA diamond",
      "score": 94.61,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 97.22,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "DTBench",
      "score": 85.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-07-16",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "WeirdML",
      "score": 82.57,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-07-16",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "LMCA",
      "score": 62.05,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-07-16",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "SimpleQA Verified",
      "score": 50.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-16",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "Chess Puzzles",
      "score": 33.71,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "SimpleBench",
      "score": 52.84,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-07-16",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 75.98,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 39.02,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "Mystery Game Puzzles",
      "score": 18.48,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "MirrorCode",
      "score": 54.44,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "PostTrainBench",
      "score": 31.48,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-16",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "Surface Evolver Bench",
      "score": 95,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-07-16",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "FrontierCode",
      "score": 44.17,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-07-16",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "Furniture Assembly",
      "score": 5.95,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "FrontierSWE",
      "score": 25.87,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-07-16",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "WeirdML v3",
      "score": 17.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-07-16",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "ARC-AGI",
      "score": 94.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-07-16",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Kimi K3",
      "model_slug": "kimi-k3",
      "developer": "Moonshot AI",
      "benchmark": "APEX-Agents",
      "score": 47.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-16",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3.1 Pro",
      "model_slug": "kimi-k3-1-pro",
      "developer": "Moonshot AI",
      "benchmark": "ProofBench",
      "score": 91,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-10-23",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3.1 Pro",
      "model_slug": "kimi-k3-1-pro",
      "developer": "Moonshot AI",
      "benchmark": "GPQA diamond",
      "score": 96.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-23",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3.1 Pro",
      "model_slug": "kimi-k3-1-pro",
      "developer": "Moonshot AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-23",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3.1 Pro",
      "model_slug": "kimi-k3-1-pro",
      "developer": "Moonshot AI",
      "benchmark": "DTBench",
      "score": 88.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-10-23",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3.1 Pro",
      "model_slug": "kimi-k3-1-pro",
      "developer": "Moonshot AI",
      "benchmark": "LMCA",
      "score": 67.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-10-23",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3.1 Pro",
      "model_slug": "kimi-k3-1-pro",
      "developer": "Moonshot AI",
      "benchmark": "SimpleQA Verified",
      "score": 62.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-23",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3.1 Pro",
      "model_slug": "kimi-k3-1-pro",
      "developer": "Moonshot AI",
      "benchmark": "Chess Puzzles",
      "score": 42.13,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-23",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3.1 Pro",
      "model_slug": "kimi-k3-1-pro",
      "developer": "Moonshot AI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 82.36,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-23",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3.1 Pro",
      "model_slug": "kimi-k3-1-pro",
      "developer": "Moonshot AI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 65.85,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-23",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3.1 Pro",
      "model_slug": "kimi-k3-1-pro",
      "developer": "Moonshot AI",
      "benchmark": "Mystery Game Puzzles",
      "score": 36.11,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-23",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3.1 Pro",
      "model_slug": "kimi-k3-1-pro",
      "developer": "Moonshot AI",
      "benchmark": "Furniture Assembly",
      "score": 36.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-23",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Kimi K3.1 Pro",
      "model_slug": "kimi-k3-1-pro",
      "developer": "Moonshot AI",
      "benchmark": "WeirdML v3",
      "score": 21.01,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-10-23",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Kimi K3.1 Pro",
      "model_slug": "kimi-k3-1-pro",
      "developer": "Moonshot AI",
      "benchmark": "APEX-Agents",
      "score": 65.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-23",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 93.75,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-11-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "ProofBench",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-11-09",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 95.66,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-11-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 99.72,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-11-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 98.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-11-09",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 83.07,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-11-09",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 94,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-11-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 77.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-11-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 93.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-11-09",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-11-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-11-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "EBR-bench",
      "score": 90.48,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-11-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "Mystery Game Puzzles",
      "score": 88.98,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-11-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "Remote Labor Index",
      "score": 41.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-11-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "Furniture Assembly",
      "score": 85.71,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-11-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "FrontierSWE",
      "score": 69.94,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-11-09",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "SpireClimb",
      "score": 19.62,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-11-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 98.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-11-09",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Fable 5.2",
      "model_slug": "claude-fable-5-2",
      "developer": "Anthropic",
      "benchmark": "APEX-Agents",
      "score": 80.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-11-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "LLaMA-13B",
      "model_slug": "llama-13b",
      "developer": "Meta AI",
      "benchmark": "ARC AI2",
      "score": 36.93,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-13B",
      "model_slug": "llama-13b",
      "developer": "Meta AI",
      "benchmark": "OpenBookQA",
      "score": 41.87,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-02-24",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-13B",
      "model_slug": "llama-13b",
      "developer": "Meta AI",
      "benchmark": "Winogrande",
      "score": 46,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "LLaMA: Open and Efficient Foundation Language Models",
      "measured_at": "2023-02-24",
      "source": "LLaMA: Open and Efficient Foundation Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-13B",
      "model_slug": "llama-13b",
      "developer": "Meta AI",
      "benchmark": "TriviaQA",
      "score": 77.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-02-24",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "LLaMA-13B",
      "model_slug": "llama-13b",
      "developer": "Meta AI",
      "benchmark": "BBH",
      "score": 17.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-13B",
      "model_slug": "llama-13b",
      "developer": "Meta AI",
      "benchmark": "GSM8K",
      "score": 20.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Baichuan 2: Open Large-scale Language Models",
      "measured_at": "2023-02-24",
      "source": "Baichuan 2: Open Large-scale Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-13B",
      "model_slug": "llama-13b",
      "developer": "Meta AI",
      "benchmark": "HellaSwag",
      "score": 72.27,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "LLaMA: Open and Efficient Foundation Language Models",
      "measured_at": "2023-02-24",
      "source": "LLaMA: Open and Efficient Foundation Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-13B",
      "model_slug": "llama-13b",
      "developer": "Meta AI",
      "benchmark": "LAMBADA",
      "score": 75.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "LLaMA-13B",
      "model_slug": "llama-13b",
      "developer": "Meta AI",
      "benchmark": "MMLU",
      "score": 30.27,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-13B",
      "model_slug": "llama-13b",
      "developer": "Meta AI",
      "benchmark": "PIQA",
      "score": 60.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "LLaMA-33B",
      "model_slug": "llama-33b",
      "developer": "Meta AI",
      "benchmark": "ARC AI2",
      "score": 56.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-33B",
      "model_slug": "llama-33b",
      "developer": "Meta AI",
      "benchmark": "OpenBookQA",
      "score": 44.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-02-24",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-33B",
      "model_slug": "llama-33b",
      "developer": "Meta AI",
      "benchmark": "Winogrande",
      "score": 52,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "LLaMA: Open and Efficient Foundation Language Models",
      "measured_at": "2023-02-24",
      "source": "LLaMA: Open and Efficient Foundation Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-33B",
      "model_slug": "llama-33b",
      "developer": "Meta AI",
      "benchmark": "TriviaQA",
      "score": 83.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-02-24",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "LLaMA-33B",
      "model_slug": "llama-33b",
      "developer": "Meta AI",
      "benchmark": "BBH",
      "score": 33.33,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-33B",
      "model_slug": "llama-33b",
      "developer": "Meta AI",
      "benchmark": "GSM8K",
      "score": 44.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Mixtral of Experts",
      "measured_at": "2023-02-24",
      "source": "Mixtral of Experts",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-33B",
      "model_slug": "llama-33b",
      "developer": "Meta AI",
      "benchmark": "HellaSwag",
      "score": 77.07,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "LLaMA: Open and Efficient Foundation Language Models",
      "measured_at": "2023-02-24",
      "source": "LLaMA: Open and Efficient Foundation Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-33B",
      "model_slug": "llama-33b",
      "developer": "Meta AI",
      "benchmark": "LAMBADA",
      "score": 77.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "LLaMA-33B",
      "model_slug": "llama-33b",
      "developer": "Meta AI",
      "benchmark": "MMLU",
      "score": 44.93,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-33B",
      "model_slug": "llama-33b",
      "developer": "Meta AI",
      "benchmark": "PIQA",
      "score": 64.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "LLaMA-65B",
      "model_slug": "llama-65b",
      "developer": "Meta AI",
      "benchmark": "ARC AI2",
      "score": 59.33,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-65B",
      "model_slug": "llama-65b",
      "developer": "Meta AI",
      "benchmark": "OpenBookQA",
      "score": 46.93,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-02-24",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-65B",
      "model_slug": "llama-65b",
      "developer": "Meta AI",
      "benchmark": "Winogrande",
      "score": 54,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "LLaMA: Open and Efficient Foundation Language Models",
      "measured_at": "2023-02-24",
      "source": "LLaMA: Open and Efficient Foundation Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-65B",
      "model_slug": "llama-65b",
      "developer": "Meta AI",
      "benchmark": "TriviaQA",
      "score": 86,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-02-24",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "LLaMA-65B",
      "model_slug": "llama-65b",
      "developer": "Meta AI",
      "benchmark": "BBH",
      "score": 44.53,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-65B",
      "model_slug": "llama-65b",
      "developer": "Meta AI",
      "benchmark": "GSM8K",
      "score": 54.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-65B",
      "model_slug": "llama-65b",
      "developer": "Meta AI",
      "benchmark": "HellaSwag",
      "score": 78.93,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "LLaMA: Open and Efficient Foundation Language Models",
      "measured_at": "2023-02-24",
      "source": "LLaMA: Open and Efficient Foundation Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-65B",
      "model_slug": "llama-65b",
      "developer": "Meta AI",
      "benchmark": "LAMBADA",
      "score": 77.7,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "LLaMA-65B",
      "model_slug": "llama-65b",
      "developer": "Meta AI",
      "benchmark": "MMLU",
      "score": 51.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-02-24",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-65B",
      "model_slug": "llama-65b",
      "developer": "Meta AI",
      "benchmark": "PIQA",
      "score": 65.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "LLaMA-7B",
      "model_slug": "llama-7b",
      "developer": "Meta AI",
      "benchmark": "ARC AI2",
      "score": 30.13,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-7B",
      "model_slug": "llama-7b",
      "developer": "Meta AI",
      "benchmark": "OpenBookQA",
      "score": 42.93,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-02-24",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-7B",
      "model_slug": "llama-7b",
      "developer": "Meta AI",
      "benchmark": "Winogrande",
      "score": 40.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "LLaMA: Open and Efficient Foundation Language Models",
      "measured_at": "2023-02-24",
      "source": "LLaMA: Open and Efficient Foundation Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-7B",
      "model_slug": "llama-7b",
      "developer": "Meta AI",
      "benchmark": "TriviaQA",
      "score": 71,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-02-24",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "LLaMA-7B",
      "model_slug": "llama-7b",
      "developer": "Meta AI",
      "benchmark": "BBH",
      "score": 11.33,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-7B",
      "model_slug": "llama-7b",
      "developer": "Meta AI",
      "benchmark": "GSM8K",
      "score": 11,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-02-24",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-7B",
      "model_slug": "llama-7b",
      "developer": "Meta AI",
      "benchmark": "HellaSwag",
      "score": 68.27,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-02-24",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-7B",
      "model_slug": "llama-7b",
      "developer": "Meta AI",
      "benchmark": "LAMBADA",
      "score": 73.3,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "LLaMA-7B",
      "model_slug": "llama-7b",
      "developer": "Meta AI",
      "benchmark": "MMLU",
      "score": 14.13,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "LLaMA-7B",
      "model_slug": "llama-7b",
      "developer": "Meta AI",
      "benchmark": "PIQA",
      "score": 59.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-02-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 2-13B",
      "model_slug": "llama-2-13b",
      "developer": "Meta AI",
      "benchmark": "ARC AI2",
      "score": 47.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-18",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-13B",
      "model_slug": "llama-2-13b",
      "developer": "Meta AI",
      "benchmark": "OpenBookQA",
      "score": 42.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-07-18",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-13B",
      "model_slug": "llama-2-13b",
      "developer": "Meta AI",
      "benchmark": "Winogrande",
      "score": 45.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-07-18",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-13B",
      "model_slug": "llama-2-13b",
      "developer": "Meta AI",
      "benchmark": "TriviaQA",
      "score": 79.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2023-07-18",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 2-13B",
      "model_slug": "llama-2-13b",
      "developer": "Meta AI",
      "benchmark": "DTBench",
      "score": 3.73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2023-07-18",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 2-13B",
      "model_slug": "llama-2-13b",
      "developer": "Meta AI",
      "benchmark": "BBH",
      "score": 44.27,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2023-07-18",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-13B",
      "model_slug": "llama-2-13b",
      "developer": "Meta AI",
      "benchmark": "GSM8K",
      "score": 36.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2023-07-18",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-13B",
      "model_slug": "llama-2-13b",
      "developer": "Meta AI",
      "benchmark": "HellaSwag",
      "score": 74.27,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemma: Open Models Based on Gemini Research and Technology",
      "measured_at": "2023-07-18",
      "source": "Gemma: Open Models Based on Gemini Research and Technology",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-13B",
      "model_slug": "llama-2-13b",
      "developer": "Meta AI",
      "benchmark": "LAMBADA",
      "score": 76.5,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-18",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 2-13B",
      "model_slug": "llama-2-13b",
      "developer": "Meta AI",
      "benchmark": "MMLU",
      "score": 40.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Mixtral of Experts",
      "measured_at": "2023-07-18",
      "source": "Mixtral of Experts",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-13B",
      "model_slug": "llama-2-13b",
      "developer": "Meta AI",
      "benchmark": "PIQA",
      "score": 61.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Mixtral of Experts",
      "measured_at": "2023-07-18",
      "source": "Mixtral of Experts",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 2-34B",
      "model_slug": "llama-2-34b",
      "developer": "Meta AI",
      "benchmark": "ARC AI2",
      "score": 39.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-07-18",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-34B",
      "model_slug": "llama-2-34b",
      "developer": "Meta AI",
      "benchmark": "OpenBookQA",
      "score": 44.27,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-07-18",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-34B",
      "model_slug": "llama-2-34b",
      "developer": "Meta AI",
      "benchmark": "Winogrande",
      "score": 53.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-07-18",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-34B",
      "model_slug": "llama-2-34b",
      "developer": "Meta AI",
      "benchmark": "TriviaQA",
      "score": 84.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-07-18",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 2-34B",
      "model_slug": "llama-2-34b",
      "developer": "Meta AI",
      "benchmark": "BBH",
      "score": 25.47,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Nemotron-4 15B Technical Report",
      "measured_at": "2023-07-18",
      "source": "Nemotron-4 15B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-34B",
      "model_slug": "llama-2-34b",
      "developer": "Meta AI",
      "benchmark": "GSM8K",
      "score": 42.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Nemotron-4 15B Technical Report",
      "measured_at": "2023-07-18",
      "source": "Nemotron-4 15B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-34B",
      "model_slug": "llama-2-34b",
      "developer": "Meta AI",
      "benchmark": "MMLU",
      "score": 50.13,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-07-18",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-34B",
      "model_slug": "llama-2-34b",
      "developer": "Meta AI",
      "benchmark": "PIQA",
      "score": 63.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-07-18",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 2-70B",
      "model_slug": "llama-2-70b",
      "developer": "Meta AI",
      "benchmark": "ARC AI2",
      "score": 71.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-18",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-70B",
      "model_slug": "llama-2-70b",
      "developer": "Meta AI",
      "benchmark": "OpenBookQA",
      "score": 46.93,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-07-18",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-70B",
      "model_slug": "llama-2-70b",
      "developer": "Meta AI",
      "benchmark": "Winogrande",
      "score": 60.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-07-18",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-70B",
      "model_slug": "llama-2-70b",
      "developer": "Meta AI",
      "benchmark": "GPQA diamond",
      "score": 1.84,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2023-07-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 2-70B",
      "model_slug": "llama-2-70b",
      "developer": "Meta AI",
      "benchmark": "MATH level 5",
      "score": 3.29,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2023-07-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 2-70B",
      "model_slug": "llama-2-70b",
      "developer": "Meta AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2023-07-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 2-70B",
      "model_slug": "llama-2-70b",
      "developer": "Meta AI",
      "benchmark": "TriviaQA",
      "score": 87.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-07-18",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 2-70B",
      "model_slug": "llama-2-70b",
      "developer": "Meta AI",
      "benchmark": "DTBench",
      "score": 2.75,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2023-07-18",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 2-70B",
      "model_slug": "llama-2-70b",
      "developer": "Meta AI",
      "benchmark": "BBH",
      "score": 53.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-18",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-70B",
      "model_slug": "llama-2-70b",
      "developer": "Meta AI",
      "benchmark": "GSM8K",
      "score": 69.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Mixtral of Experts",
      "measured_at": "2023-07-18",
      "source": "Mixtral of Experts",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-70B",
      "model_slug": "llama-2-70b",
      "developer": "Meta AI",
      "benchmark": "HellaSwag",
      "score": 80.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-18",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-70B",
      "model_slug": "llama-2-70b",
      "developer": "Meta AI",
      "benchmark": "LAMBADA",
      "score": 78.9,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-18",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 2-70B",
      "model_slug": "llama-2-70b",
      "developer": "Meta AI",
      "benchmark": "MMLU",
      "score": 59.87,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Mixtral of Experts",
      "measured_at": "2023-07-18",
      "source": "Mixtral of Experts",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 2-70B",
      "model_slug": "llama-2-70b",
      "developer": "Meta AI",
      "benchmark": "PIQA",
      "score": 65.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-18",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 3-70B",
      "model_slug": "llama-3-70b",
      "developer": "Meta AI",
      "benchmark": "Winogrande",
      "score": 67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Llama 3 Herd of Models",
      "measured_at": "2024-04-18",
      "source": "The Llama 3 Herd of Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3-70B",
      "model_slug": "llama-3-70b",
      "developer": "Meta AI",
      "benchmark": "GPQA diamond",
      "score": 21.61,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-04-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3-70B",
      "model_slug": "llama-3-70b",
      "developer": "Meta AI",
      "benchmark": "MATH level 5",
      "score": 22.55,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-04-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3-70B",
      "model_slug": "llama-3-70b",
      "developer": "Meta AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 4.21,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-04-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3-70B",
      "model_slug": "llama-3-70b",
      "developer": "Meta AI",
      "benchmark": "DTBench",
      "score": 23.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-04-18",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 3-70B",
      "model_slug": "llama-3-70b",
      "developer": "Meta AI",
      "benchmark": "Cybench",
      "score": 5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Cybench leaderboard",
      "measured_at": "2024-04-18",
      "source": "Cybench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3-70B",
      "model_slug": "llama-3-70b",
      "developer": "Meta AI",
      "benchmark": "MMLU",
      "score": 72.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-04-18",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3-8B",
      "model_slug": "llama-3-8b",
      "developer": "Meta AI",
      "benchmark": "ARC AI2",
      "score": 77.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-18",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3-8B",
      "model_slug": "llama-3-8b",
      "developer": "Meta AI",
      "benchmark": "OpenBookQA",
      "score": 76.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-18",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3-8B",
      "model_slug": "llama-3-8b",
      "developer": "Meta AI",
      "benchmark": "Winogrande",
      "score": 51.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Llama 3 Herd of Models",
      "measured_at": "2024-04-18",
      "source": "The Llama 3 Herd of Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3-8B",
      "model_slug": "llama-3-8b",
      "developer": "Meta AI",
      "benchmark": "GPQA diamond",
      "score": 1.49,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-04-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3-8B",
      "model_slug": "llama-3-8b",
      "developer": "Meta AI",
      "benchmark": "MATH level 5",
      "score": 6.13,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-04-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3-8B",
      "model_slug": "llama-3-8b",
      "developer": "Meta AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 1.85,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-04-18",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3-8B",
      "model_slug": "llama-3-8b",
      "developer": "Meta AI",
      "benchmark": "TriviaQA",
      "score": 67.7,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-18",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 3-8B",
      "model_slug": "llama-3-8b",
      "developer": "Meta AI",
      "benchmark": "DTBench",
      "score": 6.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-04-18",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 3-8B",
      "model_slug": "llama-3-8b",
      "developer": "Meta AI",
      "benchmark": "ANLI",
      "score": 35.95,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-18",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 3-8B",
      "model_slug": "llama-3-8b",
      "developer": "Meta AI",
      "benchmark": "MMLU",
      "score": 58.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-04-18",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 4.03,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-10-15",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 64.18,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-10-15",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "MATH level 5",
      "score": 96.36,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-10-15",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 66.63,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-10-15",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 56,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-10-15",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 45.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-10-15",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 36.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-10-15",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 13.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-10-15",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 3.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-10-15",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 10.36,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-10-15",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "Balrog",
      "score": 31.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Balrog Leaderboard",
      "measured_at": "2025-10-15",
      "source": "Balrog Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 3.47,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-10-15",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "DeepResearch Bench",
      "score": 45.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2025-10-15",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "Terminal Bench",
      "score": 35.51,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2025-10-15",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "ExploitBench",
      "score": 13.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-10-15",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 4.5",
      "model_slug": "claude-haiku-4-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 47.67,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-10-15",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3.1-405B",
      "model_slug": "llama-3-1-405b",
      "developer": "Meta AI",
      "benchmark": "ARC AI2",
      "score": 93.73,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-07-23",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3.1-405B",
      "model_slug": "llama-3-1-405b",
      "developer": "Meta AI",
      "benchmark": "Winogrande",
      "score": 78.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-07-23",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3.1-405B",
      "model_slug": "llama-3-1-405b",
      "developer": "Meta AI",
      "benchmark": "GPQA diamond",
      "score": 35.99,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-405B",
      "model_slug": "llama-3-1-405b",
      "developer": "Meta AI",
      "benchmark": "MATH level 5",
      "score": 49.77,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-405B",
      "model_slug": "llama-3-1-405b",
      "developer": "Meta AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 9.63,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-405B",
      "model_slug": "llama-3-1-405b",
      "developer": "Meta AI",
      "benchmark": "TriviaQA",
      "score": 82.7,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-07-23",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 3.1-405B",
      "model_slug": "llama-3-1-405b",
      "developer": "Meta AI",
      "benchmark": "DTBench",
      "score": 35.63,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-07-23",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 3.1-405B",
      "model_slug": "llama-3-1-405b",
      "developer": "Meta AI",
      "benchmark": "BBH",
      "score": 77.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-07-23",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3.1-405B",
      "model_slug": "llama-3-1-405b",
      "developer": "Meta AI",
      "benchmark": "WeirdML",
      "score": 21.38,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-07-23",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-405B",
      "model_slug": "llama-3-1-405b",
      "developer": "Meta AI",
      "benchmark": "Cybench",
      "score": 7.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Cybench leaderboard",
      "measured_at": "2024-07-23",
      "source": "Cybench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-405B",
      "model_slug": "llama-3-1-405b",
      "developer": "Meta AI",
      "benchmark": "SimpleBench",
      "score": 7.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2024-07-23",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 3.1-405B",
      "model_slug": "llama-3-1-405b",
      "developer": "Meta AI",
      "benchmark": "The Agent Company",
      "score": 7.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "TheAgentCompany experiment results github",
      "measured_at": "2024-07-23",
      "source": "TheAgentCompany experiment results github",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-405B",
      "model_slug": "llama-3-1-405b",
      "developer": "Meta AI",
      "benchmark": "HellaSwag",
      "score": 85.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-07-23",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3.1-405B",
      "model_slug": "llama-3-1-405b",
      "developer": "Meta AI",
      "benchmark": "MMLU",
      "score": 79.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-07-23",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3.1-405B",
      "model_slug": "llama-3-1-405b",
      "developer": "Meta AI",
      "benchmark": "PIQA",
      "score": 71.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-07-23",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Llama 3.1-70B",
      "model_slug": "llama-3-1-70b",
      "developer": "Meta AI",
      "benchmark": "GPQA diamond",
      "score": 26.66,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-70B",
      "model_slug": "llama-3-1-70b",
      "developer": "Meta AI",
      "benchmark": "MATH level 5",
      "score": 36.68,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-70B",
      "model_slug": "llama-3-1-70b",
      "developer": "Meta AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 3.51,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-70B",
      "model_slug": "llama-3-1-70b",
      "developer": "Meta AI",
      "benchmark": "DTBench",
      "score": 33.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-07-23",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 3.1-70B",
      "model_slug": "llama-3-1-70b",
      "developer": "Meta AI",
      "benchmark": "WeirdML",
      "score": 8.97,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-07-23",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-70B",
      "model_slug": "llama-3-1-70b",
      "developer": "Meta AI",
      "benchmark": "LMCA",
      "score": 17.47,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2024-07-23",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 3.1-70B",
      "model_slug": "llama-3-1-70b",
      "developer": "Meta AI",
      "benchmark": "Balrog",
      "score": 27.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2024-07-23",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-70B",
      "model_slug": "llama-3-1-70b",
      "developer": "Meta AI",
      "benchmark": "The Agent Company",
      "score": 6.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "TheAgentCompany experiment results github",
      "measured_at": "2024-07-23",
      "source": "TheAgentCompany experiment results github",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.1-70B",
      "model_slug": "llama-3-1-70b",
      "developer": "Meta AI",
      "benchmark": "MMLU",
      "score": 73.47,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-07-23",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 3.2 90B",
      "model_slug": "llama-3-2-90b",
      "developer": "Meta AI",
      "benchmark": "GPQA diamond",
      "score": 22.27,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.2 90B",
      "model_slug": "llama-3-2-90b",
      "developer": "Meta AI",
      "benchmark": "MATH level 5",
      "score": 39.44,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.2 90B",
      "model_slug": "llama-3-2-90b",
      "developer": "Meta AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 2.54,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.2 90B",
      "model_slug": "llama-3-2-90b",
      "developer": "Meta AI",
      "benchmark": "Balrog",
      "score": 27.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2024-09-24",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.2 90B",
      "model_slug": "llama-3-2-90b",
      "developer": "Meta AI",
      "benchmark": "GeoBench",
      "score": 52,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2024-09-24",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 3.2 90B",
      "model_slug": "llama-3-2-90b",
      "developer": "Meta AI",
      "benchmark": "MMLU",
      "score": 73.73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-09-24",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 4 Maverick",
      "model_slug": "llama-4-maverick",
      "developer": "Meta AI",
      "benchmark": "ARC-AGI-2",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 4 Maverick",
      "model_slug": "llama-4-maverick",
      "developer": "Meta AI",
      "benchmark": "GPQA diamond",
      "score": 58.31,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 4 Maverick",
      "model_slug": "llama-4-maverick",
      "developer": "Meta AI",
      "benchmark": "MATH level 5",
      "score": 73.02,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 4 Maverick",
      "model_slug": "llama-4-maverick",
      "developer": "Meta AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 20.48,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 4 Maverick",
      "model_slug": "llama-4-maverick",
      "developer": "Meta AI",
      "benchmark": "DTBench",
      "score": 36.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-04-05",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 4 Maverick",
      "model_slug": "llama-4-maverick",
      "developer": "Meta AI",
      "benchmark": "WeirdML",
      "score": 24.47,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-04-05",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 4 Maverick",
      "model_slug": "llama-4-maverick",
      "developer": "Meta AI",
      "benchmark": "LMCA",
      "score": 18.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-04-05",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 4 Maverick",
      "model_slug": "llama-4-maverick",
      "developer": "Meta AI",
      "benchmark": "SimpleBench",
      "score": 13.24,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-04-05",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 4 Maverick",
      "model_slug": "llama-4-maverick",
      "developer": "Meta AI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 1.21,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 4 Maverick",
      "model_slug": "llama-4-maverick",
      "developer": "Meta AI",
      "benchmark": "Aider polyglot",
      "score": 15.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-04-05",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 4 Maverick",
      "model_slug": "llama-4-maverick",
      "developer": "Meta AI",
      "benchmark": "GeoBench",
      "score": 52,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2025-04-05",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 4 Maverick",
      "model_slug": "llama-4-maverick",
      "developer": "Meta AI",
      "benchmark": "Fiction.LiveBench",
      "score": 46.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-04-06",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 4 Maverick",
      "model_slug": "llama-4-maverick",
      "developer": "Meta AI",
      "benchmark": "ARC-AGI",
      "score": 4.38,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2025-04-05",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 4 Maverick",
      "model_slug": "llama-4-maverick",
      "developer": "Meta AI",
      "benchmark": "Lech Mazur Writing",
      "score": 62,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-04-05",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Llama 4 Scout",
      "model_slug": "llama-4-scout",
      "developer": "Meta AI",
      "benchmark": "ARC-AGI-2",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 4 Scout",
      "model_slug": "llama-4-scout",
      "developer": "Meta AI",
      "benchmark": "GPQA diamond",
      "score": 37.27,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 4 Scout",
      "model_slug": "llama-4-scout",
      "developer": "Meta AI",
      "benchmark": "MATH level 5",
      "score": 62.27,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 4 Scout",
      "model_slug": "llama-4-scout",
      "developer": "Meta AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 7.69,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 4 Scout",
      "model_slug": "llama-4-scout",
      "developer": "Meta AI",
      "benchmark": "DTBench",
      "score": 29.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-04-05",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 4 Scout",
      "model_slug": "llama-4-scout",
      "developer": "Meta AI",
      "benchmark": "LMCA",
      "score": 14.07,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-04-05",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 4 Scout",
      "model_slug": "llama-4-scout",
      "developer": "Meta AI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Llama 4 Scout",
      "model_slug": "llama-4-scout",
      "developer": "Meta AI",
      "benchmark": "Fiction.LiveBench",
      "score": 36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-04-05",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Llama 4 Scout",
      "model_slug": "llama-4-scout",
      "developer": "Meta AI",
      "benchmark": "ARC-AGI",
      "score": 0.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-04-05",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "MPT-30B",
      "model_slug": "mpt-30b",
      "developer": "MosaicML",
      "benchmark": "ARC AI2",
      "score": 34.13,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-06-22",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "MPT-30B",
      "model_slug": "mpt-30b",
      "developer": "MosaicML",
      "benchmark": "OpenBookQA",
      "score": 36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-06-22",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "MPT-30B",
      "model_slug": "mpt-30b",
      "developer": "MosaicML",
      "benchmark": "Winogrande",
      "score": 42,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-06-22",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "MPT-30B",
      "model_slug": "mpt-30b",
      "developer": "MosaicML",
      "benchmark": "TriviaQA",
      "score": 73.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-06-22",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "MPT-30B",
      "model_slug": "mpt-30b",
      "developer": "MosaicML",
      "benchmark": "BBH",
      "score": 17.33,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-06-22",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "MPT-30B",
      "model_slug": "mpt-30b",
      "developer": "MosaicML",
      "benchmark": "GSM8K",
      "score": 34.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford HELM",
      "measured_at": "2023-06-22",
      "source": "Stanford HELM",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "MPT-30B",
      "model_slug": "mpt-30b",
      "developer": "MosaicML",
      "benchmark": "MMLU",
      "score": 30.53,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-06-22",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "MPT-30B",
      "model_slug": "mpt-30b",
      "developer": "MosaicML",
      "benchmark": "PIQA",
      "score": 63.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-06-22",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "MPT-7B",
      "model_slug": "mpt-7b",
      "developer": "MosaicML",
      "benchmark": "ARC AI2",
      "score": 23.47,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-05-05",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "MPT-7B",
      "model_slug": "mpt-7b",
      "developer": "MosaicML",
      "benchmark": "OpenBookQA",
      "score": 35.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-05-05",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "MPT-7B",
      "model_slug": "mpt-7b",
      "developer": "MosaicML",
      "benchmark": "Winogrande",
      "score": 37.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-05-05",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "MPT-7B",
      "model_slug": "mpt-7b",
      "developer": "MosaicML",
      "benchmark": "TriviaQA",
      "score": 61.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-05-05",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "MPT-7B",
      "model_slug": "mpt-7b",
      "developer": "MosaicML",
      "benchmark": "BBH",
      "score": 14.13,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-05-05",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "MPT-7B",
      "model_slug": "mpt-7b",
      "developer": "MosaicML",
      "benchmark": "GSM8K",
      "score": 9.1,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-05-05",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "MPT-7B",
      "model_slug": "mpt-7b",
      "developer": "MosaicML",
      "benchmark": "HellaSwag",
      "score": 68.53,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-05-05",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "MPT-7B",
      "model_slug": "mpt-7b",
      "developer": "MosaicML",
      "benchmark": "LAMBADA",
      "score": 70,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-05-05",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "MPT-7B",
      "model_slug": "mpt-7b",
      "developer": "MosaicML",
      "benchmark": "MMLU",
      "score": 7.73,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-05-05",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "MPT-7B",
      "model_slug": "mpt-7b",
      "developer": "MosaicML",
      "benchmark": "PIQA",
      "score": 61.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-05-05",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "MiniMax-M3",
      "model_slug": "minimax-m3",
      "developer": "MiniMax",
      "benchmark": "ProofBench",
      "score": 18,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-06-01",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "MiniMax-M3",
      "model_slug": "minimax-m3",
      "developer": "MiniMax",
      "benchmark": "GPQA diamond",
      "score": 91.54,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-01",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "MiniMax-M3",
      "model_slug": "minimax-m3",
      "developer": "MiniMax",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 71.08,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-01",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "MiniMax-M3",
      "model_slug": "minimax-m3",
      "developer": "MiniMax",
      "benchmark": "DTBench",
      "score": 64.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-06-01",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "MiniMax-M3",
      "model_slug": "minimax-m3",
      "developer": "MiniMax",
      "benchmark": "LMCA",
      "score": 39.65,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-06-01",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "MiniMax-M3",
      "model_slug": "minimax-m3",
      "developer": "MiniMax",
      "benchmark": "Chess Puzzles",
      "score": 12.67,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "MiniMax-M3",
      "model_slug": "minimax-m3",
      "developer": "MiniMax",
      "benchmark": "SimpleBench",
      "score": 34.96,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-06-01",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "MiniMax-M3",
      "model_slug": "minimax-m3",
      "developer": "MiniMax",
      "benchmark": "Mystery Game Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "MiniMax-M3",
      "model_slug": "minimax-m3",
      "developer": "MiniMax",
      "benchmark": "Surface Evolver Bench",
      "score": 55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-06-01",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "MiniMax-M3",
      "model_slug": "minimax-m3",
      "developer": "MiniMax",
      "benchmark": "FrontierCode",
      "score": 14.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-06-01",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "MiniMax-M3",
      "model_slug": "minimax-m3",
      "developer": "MiniMax",
      "benchmark": "OSWorld 2.0",
      "score": 4.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://osworld-v2.xlang.ai/",
      "measured_at": "2026-06-01",
      "source": "https://osworld-v2.xlang.ai/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "MiniMax-M3",
      "model_slug": "minimax-m3",
      "developer": "MiniMax",
      "benchmark": "APEX-Agents",
      "score": 37.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-01",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 61.25,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "ProofBench",
      "score": 89,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-10-07",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 89.79,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-07",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 98.89,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-07",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 91.65,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-10-07",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 60.07,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-10-07",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 23.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-07",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 52.65,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 79.04,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 46.34,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "EBR-bench",
      "score": 43.81,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "Mystery Game Puzzles",
      "score": 22.89,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "MirrorCode",
      "score": 78.89,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "FrontierCode",
      "score": 46.36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-10-07",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "Furniture Assembly",
      "score": 25,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "FrontierSWE",
      "score": 43.77,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-10-07",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "WeirdML v3",
      "score": 26.81,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-10-07",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 89,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-10-07",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Haiku 5.5",
      "model_slug": "claude-haiku-5-5",
      "developer": "Anthropic",
      "benchmark": "APEX-Agents",
      "score": 56.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-07",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large",
      "model_slug": "mistral-large",
      "developer": "Mistral AI",
      "benchmark": "GPQA diamond",
      "score": 19.11,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-02-26",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large",
      "model_slug": "mistral-large",
      "developer": "Mistral AI",
      "benchmark": "MATH level 5",
      "score": 24.46,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-02-26",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large",
      "model_slug": "mistral-large",
      "developer": "Mistral AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 1.85,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-02-26",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large",
      "model_slug": "mistral-large",
      "developer": "Mistral AI",
      "benchmark": "DTBench",
      "score": 26.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-02-26",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Large",
      "model_slug": "mistral-large",
      "developer": "Mistral AI",
      "benchmark": "MMLU",
      "score": 58.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-02-26",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mistral Large 2 (Jul 2024)",
      "model_slug": "mistral-large-2-jul-2024",
      "developer": "Mistral AI",
      "benchmark": "GPQA diamond",
      "score": 33.36,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large 2 (Jul 2024)",
      "model_slug": "mistral-large-2-jul-2024",
      "developer": "Mistral AI",
      "benchmark": "MATH level 5",
      "score": 44.82,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large 2 (Jul 2024)",
      "model_slug": "mistral-large-2-jul-2024",
      "developer": "Mistral AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 8.38,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large 2 (Jul 2024)",
      "model_slug": "mistral-large-2-jul-2024",
      "developer": "Mistral AI",
      "benchmark": "DTBench",
      "score": 35.35,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-07-24",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Large 2 (Jul 2024)",
      "model_slug": "mistral-large-2-jul-2024",
      "developer": "Mistral AI",
      "benchmark": "SimpleBench",
      "score": 7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2024-07-24",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Large 2 (Jul 2024)",
      "model_slug": "mistral-large-2-jul-2024",
      "developer": "Mistral AI",
      "benchmark": "MMLU",
      "score": 73.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-07-24",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mistral Large 2 (Jul 2024)",
      "model_slug": "mistral-large-2-jul-2024",
      "developer": "Mistral AI",
      "benchmark": "Lech Mazur Writing",
      "score": 69,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2024-07-24",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mistral Large 4",
      "model_slug": "mistral-large-4",
      "developer": "Mistral AI",
      "benchmark": "GPQA diamond",
      "score": 90.49,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large 4",
      "model_slug": "mistral-large-4",
      "developer": "Mistral AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 99.44,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large 4",
      "model_slug": "mistral-large-4",
      "developer": "Mistral AI",
      "benchmark": "SimpleQA Verified",
      "score": 22.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large 4",
      "model_slug": "mistral-large-4",
      "developer": "Mistral AI",
      "benchmark": "Chess Puzzles",
      "score": 25.29,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-29",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Large 4",
      "model_slug": "mistral-large-4",
      "developer": "Mistral AI",
      "benchmark": "EBR-bench",
      "score": 18.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-29",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Large 4",
      "model_slug": "mistral-large-4",
      "developer": "Mistral AI",
      "benchmark": "Mystery Game Puzzles",
      "score": 11.87,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-29",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Large 4",
      "model_slug": "mistral-large-4",
      "developer": "Mistral AI",
      "benchmark": "Furniture Assembly",
      "score": 1.19,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-29",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "ProofBench",
      "score": 10,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-10-06",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "GPQA diamond",
      "score": 87.59,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-06",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 98.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-06",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "DTBench",
      "score": 80.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-10-06",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "LMCA",
      "score": 42.11,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-10-06",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "SimpleQA Verified",
      "score": 20,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-06",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "Chess Puzzles",
      "score": 29.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-06",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 61.68,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-06",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 26.83,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-06",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "Mystery Game Puzzles",
      "score": 8.57,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-06",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "MirrorCode",
      "score": 27.78,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-06",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "Surface Evolver Bench",
      "score": 28.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-10-06",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "FrontierCode",
      "score": 39.43,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-10-06",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "Furniture Assembly",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-06",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "WeirdML v3",
      "score": 13.56,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-10-06",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Large 4 Preview",
      "model_slug": "mistral-large-4-preview",
      "developer": "Mistral AI",
      "benchmark": "APEX-Agents",
      "score": 51.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-06",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Medium 3",
      "model_slug": "mistral-medium-3",
      "developer": "Mistral AI",
      "benchmark": "GPQA diamond",
      "score": 47.96,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Medium 3",
      "model_slug": "mistral-medium-3",
      "developer": "Mistral AI",
      "benchmark": "MATH level 5",
      "score": 81.63,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Medium 3",
      "model_slug": "mistral-medium-3",
      "developer": "Mistral AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 32.15,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Medium 3",
      "model_slug": "mistral-medium-3",
      "developer": "Mistral AI",
      "benchmark": "DTBench",
      "score": 37.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-05-07",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Medium 3",
      "model_slug": "mistral-medium-3",
      "developer": "Mistral AI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 0.61,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Medium 3",
      "model_slug": "mistral-medium-3",
      "developer": "Mistral AI",
      "benchmark": "Lech Mazur Writing",
      "score": 77.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-05-07",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mistral Medium 3.5",
      "model_slug": "mistral-medium-3-5",
      "developer": "Mistral AI",
      "benchmark": "ProofBench",
      "score": 9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-04-28",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Medium 3.5",
      "model_slug": "mistral-medium-3-5",
      "developer": "Mistral AI",
      "benchmark": "DTBench",
      "score": 59.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-04-28",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Medium 3.5",
      "model_slug": "mistral-medium-3-5",
      "developer": "Mistral AI",
      "benchmark": "WeirdML",
      "score": 43.71,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-04-28",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Medium 3.5",
      "model_slug": "mistral-medium-3-5",
      "developer": "Mistral AI",
      "benchmark": "LMCA",
      "score": 30.66,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-04-28",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral Medium 3.5",
      "model_slug": "mistral-medium-3-5",
      "developer": "Mistral AI",
      "benchmark": "Surface Evolver Bench",
      "score": 26.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-04-28",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral Medium 3.5",
      "model_slug": "mistral-medium-3-5",
      "developer": "Mistral AI",
      "benchmark": "FrontierCode",
      "score": 8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-04-28",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral NeMo",
      "model_slug": "mistral-nemo",
      "developer": "Mistral AI",
      "benchmark": "GPQA diamond",
      "score": 6.8,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral NeMo",
      "model_slug": "mistral-nemo",
      "developer": "Mistral AI",
      "benchmark": "MATH level 5",
      "score": 10.83,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral NeMo",
      "model_slug": "mistral-nemo",
      "developer": "Mistral AI",
      "benchmark": "DTBench",
      "score": 14.35,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-07-18",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mistral NeMo",
      "model_slug": "mistral-nemo",
      "developer": "Mistral AI",
      "benchmark": "Balrog",
      "score": 17.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2024-07-18",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mistral NeMo",
      "model_slug": "mistral-nemo",
      "developer": "Mistral AI",
      "benchmark": "GSM8K",
      "score": 84.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-07-18",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mistral NeMo",
      "model_slug": "mistral-nemo",
      "developer": "Mistral AI",
      "benchmark": "PIQA",
      "score": 67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-07-18",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Mixtral 8x22B",
      "model_slug": "mixtral-8x22b",
      "developer": "Mistral AI",
      "benchmark": "GPQA diamond",
      "score": 12.58,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-04-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mixtral 8x22B",
      "model_slug": "mixtral-8x22b",
      "developer": "Mistral AI",
      "benchmark": "MATH level 5",
      "score": 24.24,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-04-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mixtral 8x22B",
      "model_slug": "mixtral-8x22b",
      "developer": "Mistral AI",
      "benchmark": "DTBench",
      "score": 25.23,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-04-17",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mixtral 8x22B",
      "model_slug": "mixtral-8x22b",
      "developer": "Mistral AI",
      "benchmark": "WeirdML",
      "score": 3.17,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-04-17",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mixtral 8x22B",
      "model_slug": "mixtral-8x22b",
      "developer": "Mistral AI",
      "benchmark": "Cybench",
      "score": 7.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Cybench leaderboard",
      "measured_at": "2024-04-17",
      "source": "Cybench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mixtral 8x22B",
      "model_slug": "mixtral-8x22b",
      "developer": "Mistral AI",
      "benchmark": "MMLU",
      "score": 70.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-04-17",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mixtral 8x7B",
      "model_slug": "mixtral-8x7b",
      "developer": "Mistral AI",
      "benchmark": "ARC AI2",
      "score": 83.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-12-11",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mixtral 8x7B",
      "model_slug": "mixtral-8x7b",
      "developer": "Mistral AI",
      "benchmark": "OpenBookQA",
      "score": 81.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-12-11",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mixtral 8x7B",
      "model_slug": "mixtral-8x7b",
      "developer": "Mistral AI",
      "benchmark": "Winogrande",
      "score": 54.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Mixtral of Experts",
      "measured_at": "2023-12-11",
      "source": "Mixtral of Experts",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mixtral 8x7B",
      "model_slug": "mixtral-8x7b",
      "developer": "Mistral AI",
      "benchmark": "GPQA diamond",
      "score": 7.76,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2023-12-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mixtral 8x7B",
      "model_slug": "mixtral-8x7b",
      "developer": "Mistral AI",
      "benchmark": "MATH level 5",
      "score": 9.95,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2023-12-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Mixtral 8x7B",
      "model_slug": "mixtral-8x7b",
      "developer": "Mistral AI",
      "benchmark": "TriviaQA",
      "score": 82.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-12-11",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Mixtral 8x7B",
      "model_slug": "mixtral-8x7b",
      "developer": "Mistral AI",
      "benchmark": "DTBench",
      "score": 16.02,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2023-12-11",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Mixtral 8x7B",
      "model_slug": "mixtral-8x7b",
      "developer": "Mistral AI",
      "benchmark": "GSM8K",
      "score": 74.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Mixtral of Experts",
      "measured_at": "2023-12-11",
      "source": "Mixtral of Experts",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mixtral 8x7B",
      "model_slug": "mixtral-8x7b",
      "developer": "Mistral AI",
      "benchmark": "HellaSwag",
      "score": 82.27,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Mixtral of Experts",
      "measured_at": "2023-12-11",
      "source": "Mixtral of Experts",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mixtral 8x7B",
      "model_slug": "mixtral-8x7b",
      "developer": "Mistral AI",
      "benchmark": "ANLI",
      "score": 32.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-12-11",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Mixtral 8x7B",
      "model_slug": "mixtral-8x7b",
      "developer": "Mistral AI",
      "benchmark": "MMLU",
      "score": 60.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Mixtral of Experts",
      "measured_at": "2023-12-11",
      "source": "Mixtral of Experts",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Mixtral 8x7B",
      "model_slug": "mixtral-8x7b",
      "developer": "Mistral AI",
      "benchmark": "PIQA",
      "score": 67.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Mixtral of Experts",
      "measured_at": "2023-12-11",
      "source": "Mixtral of Experts",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Claude Instant",
      "model_slug": "claude-instant",
      "developer": "Anthropic",
      "benchmark": "ARC AI2",
      "score": 81.73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Releasing Claude Instant 1.2",
      "measured_at": "2023-08-09",
      "source": "Releasing Claude Instant 1.2",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Instant",
      "model_slug": "claude-instant",
      "developer": "Anthropic",
      "benchmark": "TriviaQA",
      "score": 78.9,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Model Card and Evaluations for Claude Models",
      "measured_at": "2023-08-09",
      "source": "Model Card and Evaluations for Claude Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Claude Instant",
      "model_slug": "claude-instant",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 9.63,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2023-08-09",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Instant",
      "model_slug": "claude-instant",
      "developer": "Anthropic",
      "benchmark": "GSM8K",
      "score": 86.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Releasing Claude Instant 1.2",
      "measured_at": "2023-08-09",
      "source": "Releasing Claude Instant 1.2",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Instant",
      "model_slug": "claude-instant",
      "developer": "Anthropic",
      "benchmark": "MMLU",
      "score": 64.53,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Model Card and Evaluations for Claude Models",
      "measured_at": "2023-08-09",
      "source": "Model Card and Evaluations for Claude Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Muse Spark",
      "model_slug": "muse-spark",
      "developer": "Meta AI",
      "benchmark": "ProofBench",
      "score": 17,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-04-08",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark",
      "model_slug": "muse-spark",
      "developer": "Meta AI",
      "benchmark": "GPQA diamond",
      "score": 90,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-08",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark",
      "model_slug": "muse-spark",
      "developer": "Meta AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 88.89,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-08",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark",
      "model_slug": "muse-spark",
      "developer": "Meta AI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 68.42,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-08",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark",
      "model_slug": "muse-spark",
      "developer": "Meta AI",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 24.33,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-08",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark 1.1",
      "model_slug": "muse-spark-1-1",
      "developer": "Meta AI",
      "benchmark": "ProofBench",
      "score": 39,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-07-09",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark 1.1",
      "model_slug": "muse-spark-1-1",
      "developer": "Meta AI",
      "benchmark": "DTBench",
      "score": 90.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-07-09",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark 1.1",
      "model_slug": "muse-spark-1-1",
      "developer": "Meta AI",
      "benchmark": "LMCA",
      "score": 58.72,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-07-09",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark 1.1",
      "model_slug": "muse-spark-1-1",
      "developer": "Meta AI",
      "benchmark": "SimpleQA Verified",
      "score": 57.79,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark 1.1",
      "model_slug": "muse-spark-1-1",
      "developer": "Meta AI",
      "benchmark": "Chess Puzzles",
      "score": 30.56,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark 1.1",
      "model_slug": "muse-spark-1-1",
      "developer": "Meta AI",
      "benchmark": "Surface Evolver Bench",
      "score": 52.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-07-09",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark 1.1",
      "model_slug": "muse-spark-1-1",
      "developer": "Meta AI",
      "benchmark": "APEX-Agents",
      "score": 31.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark 1.2",
      "model_slug": "muse-spark-1-2",
      "developer": "Meta AI",
      "benchmark": "ProofBench",
      "score": 43,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-08-05",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark 1.2",
      "model_slug": "muse-spark-1-2",
      "developer": "Meta AI",
      "benchmark": "DTBench",
      "score": 91.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-08-05",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark 1.2",
      "model_slug": "muse-spark-1-2",
      "developer": "Meta AI",
      "benchmark": "WeirdML",
      "score": 60.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-08-05",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark 1.2",
      "model_slug": "muse-spark-1-2",
      "developer": "Meta AI",
      "benchmark": "LMCA",
      "score": 56.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-08-05",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark 1.2",
      "model_slug": "muse-spark-1-2",
      "developer": "Meta AI",
      "benchmark": "SimpleQA Verified",
      "score": 60.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark 1.2",
      "model_slug": "muse-spark-1-2",
      "developer": "Meta AI",
      "benchmark": "Chess Puzzles",
      "score": 23.19,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark 1.2",
      "model_slug": "muse-spark-1-2",
      "developer": "Meta AI",
      "benchmark": "SimpleBench",
      "score": 69.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-08-05",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark 1.2",
      "model_slug": "muse-spark-1-2",
      "developer": "Meta AI",
      "benchmark": "FrontierSWE",
      "score": 11.97,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-08-05",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark 1.2",
      "model_slug": "muse-spark-1-2",
      "developer": "Meta AI",
      "benchmark": "APEX-Agents",
      "score": 45.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark 1.3",
      "model_slug": "muse-spark-1-3",
      "developer": "Meta AI",
      "benchmark": "ProofBench",
      "score": 58,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-09-02",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark 1.3",
      "model_slug": "muse-spark-1-3",
      "developer": "Meta AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 99.17,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark 1.3",
      "model_slug": "muse-spark-1-3",
      "developer": "Meta AI",
      "benchmark": "DTBench",
      "score": 94.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-09-02",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark 1.3",
      "model_slug": "muse-spark-1-3",
      "developer": "Meta AI",
      "benchmark": "LMCA",
      "score": 63.38,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-09-02",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark 1.3",
      "model_slug": "muse-spark-1-3",
      "developer": "Meta AI",
      "benchmark": "Chess Puzzles",
      "score": 33.71,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark 1.3",
      "model_slug": "muse-spark-1-3",
      "developer": "Meta AI",
      "benchmark": "SimpleBench",
      "score": 78.16,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-09-02",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark 1.3",
      "model_slug": "muse-spark-1-3",
      "developer": "Meta AI",
      "benchmark": "GeoBench",
      "score": 86,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2026-09-02",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark 1.3",
      "model_slug": "muse-spark-1-3",
      "developer": "Meta AI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 78.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark 1.3",
      "model_slug": "muse-spark-1-3",
      "developer": "Meta AI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 46.34,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark 1.3",
      "model_slug": "muse-spark-1-3",
      "developer": "Meta AI",
      "benchmark": "Mystery Game Puzzles",
      "score": 17.38,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark 1.3",
      "model_slug": "muse-spark-1-3",
      "developer": "Meta AI",
      "benchmark": "MirrorCode",
      "score": 54.44,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark 1.3",
      "model_slug": "muse-spark-1-3",
      "developer": "Meta AI",
      "benchmark": "FrontierSWE",
      "score": 25.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-09-02",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark 1.3",
      "model_slug": "muse-spark-1-3",
      "developer": "Meta AI",
      "benchmark": "WeirdML v3",
      "score": 13.48,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-09-02",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark 1.3",
      "model_slug": "muse-spark-1-3",
      "developer": "Meta AI",
      "benchmark": "APEX-Agents",
      "score": 58.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-02",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark Open",
      "model_slug": "muse-spark-open",
      "developer": "Meta AI",
      "benchmark": "ARC-AGI-2",
      "score": 57.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-15",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark Open",
      "model_slug": "muse-spark-open",
      "developer": "Meta AI",
      "benchmark": "GPQA diamond",
      "score": 88.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-15",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark Open",
      "model_slug": "muse-spark-open",
      "developer": "Meta AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 98.05,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-15",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark Open",
      "model_slug": "muse-spark-open",
      "developer": "Meta AI",
      "benchmark": "SimpleQA Verified",
      "score": 56.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-15",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Spark Open",
      "model_slug": "muse-spark-open",
      "developer": "Meta AI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 73.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-15",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark Open",
      "model_slug": "muse-spark-open",
      "developer": "Meta AI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 31.71,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-15",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Spark Open",
      "model_slug": "muse-spark-open",
      "developer": "Meta AI",
      "benchmark": "ARC-AGI",
      "score": 89,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-10-15",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Muse Spark Open",
      "model_slug": "muse-spark-open",
      "developer": "Meta AI",
      "benchmark": "APEX-Agents",
      "score": 33.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-15",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Wonder",
      "model_slug": "muse-wonder",
      "developer": null,
      "benchmark": "ProofBench",
      "score": 84,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-10-12",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Wonder",
      "model_slug": "muse-wonder",
      "developer": null,
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 99.44,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-12",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Wonder",
      "model_slug": "muse-wonder",
      "developer": null,
      "benchmark": "DTBench",
      "score": 92,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-10-12",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Wonder",
      "model_slug": "muse-wonder",
      "developer": null,
      "benchmark": "LMCA",
      "score": 66.87,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-10-12",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Wonder",
      "model_slug": "muse-wonder",
      "developer": null,
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 86.43,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-12",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Wonder",
      "model_slug": "muse-wonder",
      "developer": null,
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 70.73,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-12",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Wonder",
      "model_slug": "muse-wonder",
      "developer": null,
      "benchmark": "EBR-bench",
      "score": 31.43,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-12",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Wonder",
      "model_slug": "muse-wonder",
      "developer": null,
      "benchmark": "Mystery Game Puzzles",
      "score": 46.02,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-12",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Wonder",
      "model_slug": "muse-wonder",
      "developer": null,
      "benchmark": "MirrorCode",
      "score": 45.56,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-12",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Wonder",
      "model_slug": "muse-wonder",
      "developer": null,
      "benchmark": "PostTrainBench",
      "score": 24.19,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-12",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Muse Wonder",
      "model_slug": "muse-wonder",
      "developer": null,
      "benchmark": "Furniture Assembly",
      "score": 34.52,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-12",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Muse Wonder",
      "model_slug": "muse-wonder",
      "developer": null,
      "benchmark": "WeirdML v3",
      "score": 27.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-10-12",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Muse Wonder",
      "model_slug": "muse-wonder",
      "developer": null,
      "benchmark": "APEX-Agents",
      "score": 66.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-12",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Nemotron 3 Ultra",
      "model_slug": "nemotron-3-ultra",
      "developer": "NVIDIA",
      "benchmark": "ProofBench",
      "score": 2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-06-04",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Nemotron 3 Ultra",
      "model_slug": "nemotron-3-ultra",
      "developer": "NVIDIA",
      "benchmark": "GPQA diamond",
      "score": 83.82,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-04",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Nemotron 3 Ultra",
      "model_slug": "nemotron-3-ultra",
      "developer": "NVIDIA",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 86.65,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-04",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Nemotron 3 Ultra",
      "model_slug": "nemotron-3-ultra",
      "developer": "NVIDIA",
      "benchmark": "DTBench",
      "score": 83.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-06-04",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Nemotron 3 Ultra",
      "model_slug": "nemotron-3-ultra",
      "developer": "NVIDIA",
      "benchmark": "WeirdML",
      "score": 43.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-06-04",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Nemotron 3 Ultra",
      "model_slug": "nemotron-3-ultra",
      "developer": "NVIDIA",
      "benchmark": "LMCA",
      "score": 43.41,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-06-04",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Nemotron 3 Ultra",
      "model_slug": "nemotron-3-ultra",
      "developer": "NVIDIA",
      "benchmark": "SimpleBench",
      "score": 30.04,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-06-04",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Nemotron 3 Ultra",
      "model_slug": "nemotron-3-ultra",
      "developer": "NVIDIA",
      "benchmark": "Mystery Game Puzzles",
      "score": 11.87,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-04",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Nemotron 3 Ultra",
      "model_slug": "nemotron-3-ultra",
      "developer": "NVIDIA",
      "benchmark": "FrontierCode",
      "score": 13.61,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-06-04",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Nemotron 3 Ultra",
      "model_slug": "nemotron-3-ultra",
      "developer": "NVIDIA",
      "benchmark": "APEX-Agents",
      "score": 22.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-04",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Nemotron-4 15B",
      "model_slug": "nemotron-4-15b",
      "developer": "NVIDIA",
      "benchmark": "ARC AI2",
      "score": 40.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Nemotron-4 15B Technical Report",
      "measured_at": "2024-02-26",
      "source": "Nemotron-4 15B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Nemotron-4 15B",
      "model_slug": "nemotron-4-15b",
      "developer": "NVIDIA",
      "benchmark": "Winogrande",
      "score": 56,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Nemotron-4 15B Technical Report",
      "measured_at": "2024-02-26",
      "source": "Nemotron-4 15B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Nemotron-4 15B",
      "model_slug": "nemotron-4-15b",
      "developer": "NVIDIA",
      "benchmark": "BBH",
      "score": 44.93,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Nemotron-4 15B Technical Report",
      "measured_at": "2024-02-26",
      "source": "Nemotron-4 15B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Nemotron-4 15B",
      "model_slug": "nemotron-4-15b",
      "developer": "NVIDIA",
      "benchmark": "GSM8K",
      "score": 46,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Nemotron-4 15B Technical Report",
      "measured_at": "2024-02-26",
      "source": "Nemotron-4 15B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Nemotron-4 15B",
      "model_slug": "nemotron-4-15b",
      "developer": "NVIDIA",
      "benchmark": "HellaSwag",
      "score": 76.53,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Nemotron-4 15B Technical Report",
      "measured_at": "2024-02-26",
      "source": "Nemotron-4 15B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Nemotron-4 15B",
      "model_slug": "nemotron-4-15b",
      "developer": "NVIDIA",
      "benchmark": "MMLU",
      "score": 44.93,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Nemotron-4 15B Technical Report",
      "measured_at": "2024-02-26",
      "source": "Nemotron-4 15B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Nemotron-4 15B",
      "model_slug": "nemotron-4-15b",
      "developer": "NVIDIA",
      "benchmark": "PIQA",
      "score": 64.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Nemotron-4 15B Technical Report",
      "measured_at": "2024-02-26",
      "source": "Nemotron-4 15B Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "PaLM 2-L",
      "model_slug": "palm-2-l",
      "developer": "Google DeepMind",
      "benchmark": "ARC AI2",
      "score": 58.93,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-05-17",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "PaLM 2-L",
      "model_slug": "palm-2-l",
      "developer": "Google DeepMind",
      "benchmark": "Winogrande",
      "score": 66,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "PaLM 2 Technical Report",
      "measured_at": "2023-05-17",
      "source": "PaLM 2 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "PaLM 2-L",
      "model_slug": "palm-2-l",
      "developer": "Google DeepMind",
      "benchmark": "TriviaQA",
      "score": 86.1,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "PaLM 2 Technical Report",
      "measured_at": "2023-05-17",
      "source": "PaLM 2 Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "PaLM 2-L",
      "model_slug": "palm-2-l",
      "developer": "Google DeepMind",
      "benchmark": "HellaSwag",
      "score": 82.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "PaLM 2 Technical Report",
      "measured_at": "2023-05-17",
      "source": "PaLM 2 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "PaLM 2-M",
      "model_slug": "palm-2-m",
      "developer": "Google DeepMind",
      "benchmark": "ARC AI2",
      "score": 53.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-05-17",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "PaLM 2-M",
      "model_slug": "palm-2-m",
      "developer": "Google DeepMind",
      "benchmark": "OpenBookQA",
      "score": 43.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "PaLM 2 Technical Report",
      "measured_at": "2023-05-17",
      "source": "PaLM 2 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "PaLM 2-M",
      "model_slug": "palm-2-m",
      "developer": "Google DeepMind",
      "benchmark": "Winogrande",
      "score": 58.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "PaLM 2 Technical Report",
      "measured_at": "2023-05-17",
      "source": "PaLM 2 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "PaLM 2-M",
      "model_slug": "palm-2-m",
      "developer": "Google DeepMind",
      "benchmark": "TriviaQA",
      "score": 81.7,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "PaLM 2 Technical Report",
      "measured_at": "2023-05-17",
      "source": "PaLM 2 Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "PaLM 2-M",
      "model_slug": "palm-2-m",
      "developer": "Google DeepMind",
      "benchmark": "HellaSwag",
      "score": 78.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "PaLM 2 Technical Report",
      "measured_at": "2023-05-17",
      "source": "PaLM 2 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 8.61,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-05-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 71.2,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "MATH level 5",
      "score": 85.05,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 64.41,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 69.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-05-22",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 43.72,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-05-22",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 44.04,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-05-22",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "Cybench",
      "score": 38,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Cybench leaderboard",
      "measured_at": "2025-05-22",
      "source": "Cybench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 50.56,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-05-22",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "METR Time Horizons",
      "score": 63.93,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-05-22",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 7.86,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "Aider polyglot",
      "score": 72,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-05-22",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "GeoBench",
      "score": 49,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2025-05-22",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 6.94,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "VPCT",
      "score": 7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-05-22",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "GSO-Bench",
      "score": 6.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GSO Leaderboard",
      "measured_at": "2025-05-22",
      "source": "GSO Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "DeepResearch Bench",
      "score": 46.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2025-05-22",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "Fiction.LiveBench",
      "score": 61.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-05-22",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 35.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2025-05-22",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "Lech Mazur Writing",
      "score": 83.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-05-22",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4",
      "model_slug": "claude-opus-4",
      "developer": "Anthropic",
      "benchmark": "SWE-Bench verified",
      "score": 70.66,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "PaLM 2-S",
      "model_slug": "palm-2-s",
      "developer": "Google DeepMind",
      "benchmark": "ARC AI2",
      "score": 46.13,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-05-17",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "PaLM 2-S",
      "model_slug": "palm-2-s",
      "developer": "Google DeepMind",
      "benchmark": "OpenBookQA",
      "score": 41.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "PaLM 2 Technical Report",
      "measured_at": "2023-05-17",
      "source": "PaLM 2 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "PaLM 2-S",
      "model_slug": "palm-2-s",
      "developer": "Google DeepMind",
      "benchmark": "Winogrande",
      "score": 55.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "PaLM 2 Technical Report",
      "measured_at": "2023-05-17",
      "source": "PaLM 2 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "PaLM 2-S",
      "model_slug": "palm-2-s",
      "developer": "Google DeepMind",
      "benchmark": "TriviaQA",
      "score": 75.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "PaLM 2 Technical Report",
      "measured_at": "2023-05-17",
      "source": "PaLM 2 Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "PaLM 2-S",
      "model_slug": "palm-2-s",
      "developer": "Google DeepMind",
      "benchmark": "HellaSwag",
      "score": 76,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "PaLM 2 Technical Report",
      "measured_at": "2023-05-17",
      "source": "PaLM 2 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Phi-1.5",
      "model_slug": "phi-1-5",
      "developer": "Microsoft",
      "benchmark": "ARC AI2",
      "score": 25.87,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Textbooks Are All You Need II: phi-1.5 technical report",
      "measured_at": "2023-09-11",
      "source": "Textbooks Are All You Need II: phi-1.5 technical report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Phi-1.5",
      "model_slug": "phi-1-5",
      "developer": "Microsoft",
      "benchmark": "OpenBookQA",
      "score": 16.27,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Textbooks Are All You Need II: phi-1.5 technical report",
      "measured_at": "2023-09-11",
      "source": "Textbooks Are All You Need II: phi-1.5 technical report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Phi-1.5",
      "model_slug": "phi-1-5",
      "developer": "Microsoft",
      "benchmark": "Winogrande",
      "score": 46.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Textbooks Are All You Need II: phi-1.5 technical report",
      "measured_at": "2023-09-11",
      "source": "Textbooks Are All You Need II: phi-1.5 technical report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Phi-1.5",
      "model_slug": "phi-1-5",
      "developer": "Microsoft",
      "benchmark": "HellaSwag",
      "score": 30.13,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Textbooks Are All You Need II: phi-1.5 technical report",
      "measured_at": "2023-09-11",
      "source": "Textbooks Are All You Need II: phi-1.5 technical report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Phi-1.5",
      "model_slug": "phi-1-5",
      "developer": "Microsoft",
      "benchmark": "MMLU",
      "score": 16.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Textbooks Are All You Need II: phi-1.5 technical report",
      "measured_at": "2023-09-11",
      "source": "Textbooks Are All You Need II: phi-1.5 technical report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "QwQ-32B",
      "model_slug": "qwq-32b",
      "developer": null,
      "benchmark": "GPQA diamond",
      "score": 56.03,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-03-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "QwQ-32B",
      "model_slug": "qwq-32b",
      "developer": null,
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 59.13,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-03-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "QwQ-32B",
      "model_slug": "qwq-32b",
      "developer": null,
      "benchmark": "Aider polyglot",
      "score": 20.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-03-05",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "QwQ-32B",
      "model_slug": "qwq-32b",
      "developer": null,
      "benchmark": "Fiction.LiveBench",
      "score": 83.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-03-05",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "QwQ-32B",
      "model_slug": "qwq-32b",
      "developer": null,
      "benchmark": "Lech Mazur Writing",
      "score": 80.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-03-05",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen 3.5 Flash (hosted 35B-A3B)",
      "model_slug": "qwen-3-5-flash-hosted-35b-a3b",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 79.62,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.5 Flash (hosted 35B-A3B)",
      "model_slug": "qwen-3-5-flash-hosted-35b-a3b",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 84.43,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.5 Flash (hosted 35B-A3B)",
      "model_slug": "qwen-3-5-flash-hosted-35b-a3b",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 71.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-02-25",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.5 Flash (hosted 35B-A3B)",
      "model_slug": "qwen-3-5-flash-hosted-35b-a3b",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 34.28,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-02-25",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.5 Flash (hosted 35B-A3B)",
      "model_slug": "qwen-3-5-flash-hosted-35b-a3b",
      "developer": "Qwen",
      "benchmark": "SimpleQA Verified",
      "score": 20.32,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-25",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.5 Flash (hosted 35B-A3B)",
      "model_slug": "qwen-3-5-flash-hosted-35b-a3b",
      "developer": "Qwen",
      "benchmark": "Chess Puzzles",
      "score": 9.51,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-25",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.5 Flash (hosted 35B-A3B)",
      "model_slug": "qwen-3-5-flash-hosted-35b-a3b",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.5 Flash (hosted 35B-A3B)",
      "model_slug": "qwen-3-5-flash-hosted-35b-a3b",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 19.21,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-25",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.5 Flash (hosted 35B-A3B)",
      "model_slug": "qwen-3-5-flash-hosted-35b-a3b",
      "developer": "Qwen",
      "benchmark": "Mystery Game Puzzles",
      "score": 11.87,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-25",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.5 Plus (hosted 397B-A17B)",
      "model_slug": "qwen-3-5-plus-hosted-397b-a17b",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 83.12,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.5 Plus (hosted 397B-A17B)",
      "model_slug": "qwen-3-5-plus-hosted-397b-a17b",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 86.65,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.5 Plus (hosted 397B-A17B)",
      "model_slug": "qwen-3-5-plus-hosted-397b-a17b",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 67.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-02-16",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.5 Plus (hosted 397B-A17B)",
      "model_slug": "qwen-3-5-plus-hosted-397b-a17b",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 42.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-02-16",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.5 Plus (hosted 397B-A17B)",
      "model_slug": "qwen-3-5-plus-hosted-397b-a17b",
      "developer": "Qwen",
      "benchmark": "SimpleQA Verified",
      "score": 25.35,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-16",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.5 Plus (hosted 397B-A17B)",
      "model_slug": "qwen-3-5-plus-hosted-397b-a17b",
      "developer": "Qwen",
      "benchmark": "Chess Puzzles",
      "score": 4.25,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.5 Plus (hosted 397B-A17B)",
      "model_slug": "qwen-3-5-plus-hosted-397b-a17b",
      "developer": "Qwen",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 36.9,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.5 Plus (hosted 397B-A17B)",
      "model_slug": "qwen-3-5-plus-hosted-397b-a17b",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 3.47,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.5 Plus (hosted 397B-A17B)",
      "model_slug": "qwen-3-5-plus-hosted-397b-a17b",
      "developer": "Qwen",
      "benchmark": "Mystery Game Puzzles",
      "score": 8.57,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.5 Plus (hosted 397B-A17B)",
      "model_slug": "qwen-3-5-plus-hosted-397b-a17b",
      "developer": "Qwen",
      "benchmark": "CL-bench",
      "score": 19.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-16",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen 3.5 Plus (hosted 397B-A17B)",
      "model_slug": "qwen-3-5-plus-hosted-397b-a17b",
      "developer": "Qwen",
      "benchmark": "CL-bench Life",
      "score": 12.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-16",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 35B-A3B",
      "model_slug": "qwen-3-6-35b-a3b",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 83.12,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 35B-A3B",
      "model_slug": "qwen-3-6-35b-a3b",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 86.65,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 35B-A3B",
      "model_slug": "qwen-3-6-35b-a3b",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 56.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-04-14",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 35B-A3B",
      "model_slug": "qwen-3-6-35b-a3b",
      "developer": "Qwen",
      "benchmark": "WeirdML",
      "score": 34.49,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-04-14",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 35B-A3B",
      "model_slug": "qwen-3-6-35b-a3b",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 34.94,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-04-14",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 35B-A3B",
      "model_slug": "qwen-3-6-35b-a3b",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 21.42,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 35B-A3B",
      "model_slug": "qwen-3-6-35b-a3b",
      "developer": "Qwen",
      "benchmark": "Mystery Game Puzzles",
      "score": 14.08,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 35B-A3B",
      "model_slug": "qwen-3-6-35b-a3b",
      "developer": "Qwen",
      "benchmark": "Surface Evolver Bench",
      "score": 44.38,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-04-14",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 35B-A3B",
      "model_slug": "qwen-3-6-35b-a3b",
      "developer": "Qwen",
      "benchmark": "Terminal Bench",
      "score": 24.63,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2026-04-14",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 Flash",
      "model_slug": "qwen-3-6-flash",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 81.02,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-27",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 Flash",
      "model_slug": "qwen-3-6-flash",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 84.43,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-27",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 Flash",
      "model_slug": "qwen-3-6-flash",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 61.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-04-27",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Flash",
      "model_slug": "qwen-3-6-flash",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 36.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-04-27",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Flash",
      "model_slug": "qwen-3-6-flash",
      "developer": "Qwen",
      "benchmark": "SimpleQA Verified",
      "score": 15.92,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-27",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 Flash",
      "model_slug": "qwen-3-6-flash",
      "developer": "Qwen",
      "benchmark": "Chess Puzzles",
      "score": 7.41,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-27",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Flash",
      "model_slug": "qwen-3-6-flash",
      "developer": "Qwen",
      "benchmark": "SimpleBench",
      "score": 22.24,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-04-27",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Flash",
      "model_slug": "qwen-3-6-flash",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-27",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 Flash",
      "model_slug": "qwen-3-6-flash",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 23.64,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-27",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Flash",
      "model_slug": "qwen-3-6-flash",
      "developer": "Qwen",
      "benchmark": "Mystery Game Puzzles",
      "score": 9.67,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-27",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Max (Preview)",
      "model_slug": "qwen-3-6-max-preview",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 86.63,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 Max (Preview)",
      "model_slug": "qwen-3-6-max-preview",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 91.1,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 Max (Preview)",
      "model_slug": "qwen-3-6-max-preview",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 78.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-04-20",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Max (Preview)",
      "model_slug": "qwen-3-6-max-preview",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 50.04,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-04-20",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Max (Preview)",
      "model_slug": "qwen-3-6-max-preview",
      "developer": "Qwen",
      "benchmark": "SimpleQA Verified",
      "score": 52,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-20",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 Max (Preview)",
      "model_slug": "qwen-3-6-max-preview",
      "developer": "Qwen",
      "benchmark": "SimpleBench",
      "score": 55.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2026-04-20",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Max (Preview)",
      "model_slug": "qwen-3-6-max-preview",
      "developer": "Qwen",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 40.53,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Max (Preview)",
      "model_slug": "qwen-3-6-max-preview",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 6.94,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 Max (Preview)",
      "model_slug": "qwen-3-6-max-preview",
      "developer": "Qwen",
      "benchmark": "Mystery Game Puzzles",
      "score": 10.77,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-20",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Max (Preview)",
      "model_slug": "qwen-3-6-max-preview",
      "developer": "Qwen",
      "benchmark": "SWE-Bench verified",
      "score": 76.65,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Baichuan2-13B",
      "model_slug": "baichuan2-13b",
      "developer": "Baichuan AI",
      "benchmark": "ARC AI2",
      "score": 17.33,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-06",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Baichuan2-13B",
      "model_slug": "baichuan2-13b",
      "developer": "Baichuan AI",
      "benchmark": "BBH",
      "score": 32,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-06",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Baichuan2-13B",
      "model_slug": "baichuan2-13b",
      "developer": "Baichuan AI",
      "benchmark": "GSM8K",
      "score": 52.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Nemotron-4 15B Technical Report",
      "measured_at": "2023-09-06",
      "source": "Nemotron-4 15B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Baichuan2-13B",
      "model_slug": "baichuan2-13b",
      "developer": "Baichuan AI",
      "benchmark": "HellaSwag",
      "score": 61.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-06",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Baichuan2-13B",
      "model_slug": "baichuan2-13b",
      "developer": "Baichuan AI",
      "benchmark": "LAMBADA",
      "score": 74,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-06",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Baichuan2-13B",
      "model_slug": "baichuan2-13b",
      "developer": "Baichuan AI",
      "benchmark": "MMLU",
      "score": 45.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Nemotron-4 15B Technical Report",
      "measured_at": "2023-09-06",
      "source": "Nemotron-4 15B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Baichuan2-13B",
      "model_slug": "baichuan2-13b",
      "developer": "Baichuan AI",
      "benchmark": "PIQA",
      "score": 56.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-06",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 72.6,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 68.86,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 66.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-08-05",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 45.86,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-08-05",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 43.62,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-08-05",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "Cybench",
      "score": 42,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Cybench leaderboard",
      "measured_at": "2025-08-05",
      "source": "Cybench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 52,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-08-05",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "METR Time Horizons",
      "score": 66.81,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-08-05",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "VPCT",
      "score": 2.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-08-05",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "DeepResearch Bench",
      "score": 48.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2025-08-05",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 13.3,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 2.44,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "EBR-bench",
      "score": 7.94,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "Mystery Game Puzzles",
      "score": 12.98,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "Terminal Bench",
      "score": 38,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Terminal-Bench v2 Leaderboard",
      "measured_at": "2025-08-05",
      "source": "Terminal-Bench v2 Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "GDPval",
      "score": 43.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "Lech Mazur Writing",
      "score": 84.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-08-05",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.1",
      "model_slug": "claude-opus-4-1",
      "developer": "Anthropic",
      "benchmark": "SWE-Bench verified",
      "score": 73.35,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen 3.6 Plus",
      "model_slug": "qwen-3-6-plus",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 88.03,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-31",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 Plus",
      "model_slug": "qwen-3-6-plus",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 93.33,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-31",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 Plus",
      "model_slug": "qwen-3-6-plus",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 69.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-03-31",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Plus",
      "model_slug": "qwen-3-6-plus",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 38.93,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-03-31",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Plus",
      "model_slug": "qwen-3-6-plus",
      "developer": "Qwen",
      "benchmark": "SimpleQA Verified",
      "score": 44.14,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-31",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 Plus",
      "model_slug": "qwen-3-6-plus",
      "developer": "Qwen",
      "benchmark": "Chess Puzzles",
      "score": 11.62,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-31",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Plus",
      "model_slug": "qwen-3-6-plus",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 13.89,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-31",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.6 Plus",
      "model_slug": "qwen-3-6-plus",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 40.26,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-31",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Plus",
      "model_slug": "qwen-3-6-plus",
      "developer": "Qwen",
      "benchmark": "Mystery Game Puzzles",
      "score": 3.06,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-31",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.6 Plus",
      "model_slug": "qwen-3-6-plus",
      "developer": "Qwen",
      "benchmark": "CL-bench",
      "score": 20.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-31",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen 3.6 Plus",
      "model_slug": "qwen-3-6-plus",
      "developer": "Qwen",
      "benchmark": "SWE-Bench verified",
      "score": 57.85,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-31",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen 3.8 27B",
      "model_slug": "qwen-3-8-27b",
      "developer": "Qwen",
      "benchmark": "ARC-AGI-2",
      "score": 42.36,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.8 27B",
      "model_slug": "qwen-3-8-27b",
      "developer": "Qwen",
      "benchmark": "ProofBench",
      "score": 16,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-08-14",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.8 27B",
      "model_slug": "qwen-3-8-27b",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 80,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-08-14",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.8 27B",
      "model_slug": "qwen-3-8-27b",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 48.66,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-08-14",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.8 27B",
      "model_slug": "qwen-3-8-27b",
      "developer": "Qwen",
      "benchmark": "SimpleBench",
      "score": 52.24,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-08-14",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.8 27B",
      "model_slug": "qwen-3-8-27b",
      "developer": "Qwen",
      "benchmark": "GeoBench",
      "score": 38,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2026-08-14",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.8 27B",
      "model_slug": "qwen-3-8-27b",
      "developer": "Qwen",
      "benchmark": "Surface Evolver Bench",
      "score": 45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-08-14",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.8 27B",
      "model_slug": "qwen-3-8-27b",
      "developer": "Qwen",
      "benchmark": "ARC-AGI",
      "score": 87.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-08-14",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen 3.8 27B",
      "model_slug": "qwen-3-8-27b",
      "developer": "Qwen",
      "benchmark": "APEX-Agents",
      "score": 47.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-14",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.8 Max",
      "model_slug": "qwen-3-8-max",
      "developer": "Qwen",
      "benchmark": "ProofBench",
      "score": 58,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-08-02",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.8 Max",
      "model_slug": "qwen-3-8-max",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 94,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-02",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.8 Max",
      "model_slug": "qwen-3-8-max",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 99.44,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-02",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.8 Max",
      "model_slug": "qwen-3-8-max",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 86.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-08-02",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.8 Max",
      "model_slug": "qwen-3-8-max",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 54.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-08-02",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.8 Max",
      "model_slug": "qwen-3-8-max",
      "developer": "Qwen",
      "benchmark": "SimpleQA Verified",
      "score": 45.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-02",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.8 Max",
      "model_slug": "qwen-3-8-max",
      "developer": "Qwen",
      "benchmark": "Chess Puzzles",
      "score": 43.18,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.8 Max",
      "model_slug": "qwen-3-8-max",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 78.67,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-02",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.8 Max",
      "model_slug": "qwen-3-8-max",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 46.34,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-02",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.8 Max",
      "model_slug": "qwen-3-8-max",
      "developer": "Qwen",
      "benchmark": "Mystery Game Puzzles",
      "score": 31.7,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-02",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.8 Max",
      "model_slug": "qwen-3-8-max",
      "developer": "Qwen",
      "benchmark": "MirrorCode",
      "score": 52.22,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen 3.8 Max",
      "model_slug": "qwen-3-8-max",
      "developer": "Qwen",
      "benchmark": "FrontierSWE",
      "score": 15.81,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-08-02",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen 3.8 Max",
      "model_slug": "qwen-3-8-max",
      "developer": "Qwen",
      "benchmark": "APEX-Agents",
      "score": 63.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-02",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen-14B",
      "model_slug": "qwen-14b",
      "developer": "Qwen",
      "benchmark": "ARC AI2",
      "score": 79.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen-14B",
      "model_slug": "qwen-14b",
      "developer": "Qwen",
      "benchmark": "BBH",
      "score": 40,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2023-09-24",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen-14B",
      "model_slug": "qwen-14b",
      "developer": "Qwen",
      "benchmark": "GSM8K",
      "score": 61.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2023-09-24",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen-14B",
      "model_slug": "qwen-14b",
      "developer": "Qwen",
      "benchmark": "LAMBADA",
      "score": 71.1,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Qwen-14B",
      "model_slug": "qwen-14b",
      "developer": "Qwen",
      "benchmark": "MMLU",
      "score": 55.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Nemotron-4 15B Technical Report",
      "measured_at": "2023-09-24",
      "source": "Nemotron-4 15B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen-14B",
      "model_slug": "qwen-14b",
      "developer": "Qwen",
      "benchmark": "PIQA",
      "score": 59.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Qwen-1_8B",
      "model_slug": "qwen-1-8b",
      "developer": "Qwen",
      "benchmark": "ARC AI2",
      "score": 37.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-11-30",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen-1_8B",
      "model_slug": "qwen-1-8b",
      "developer": "Qwen",
      "benchmark": "BBH",
      "score": 4.27,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-11-30",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen-1_8B",
      "model_slug": "qwen-1-8b",
      "developer": "Qwen",
      "benchmark": "GSM8K",
      "score": 21.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-11-30",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen-1_8B",
      "model_slug": "qwen-1-8b",
      "developer": "Qwen",
      "benchmark": "LAMBADA",
      "score": 58.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-11-30",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Qwen-1_8B",
      "model_slug": "qwen-1-8b",
      "developer": "Qwen",
      "benchmark": "MMLU",
      "score": 4.27,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-11-30",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen-1_8B",
      "model_slug": "qwen-1-8b",
      "developer": "Qwen",
      "benchmark": "PIQA",
      "score": 46.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-11-30",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Qwen-7B",
      "model_slug": "qwen-7b",
      "developer": "Qwen",
      "benchmark": "ARC AI2",
      "score": 67.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-28",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen-7B",
      "model_slug": "qwen-7b",
      "developer": "Qwen",
      "benchmark": "BBH",
      "score": 26.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-28",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen-7B",
      "model_slug": "qwen-7b",
      "developer": "Qwen",
      "benchmark": "GSM8K",
      "score": 51.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2023-09-28",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen-7B",
      "model_slug": "qwen-7b",
      "developer": "Qwen",
      "benchmark": "LAMBADA",
      "score": 67.9,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-28",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Qwen-7B",
      "model_slug": "qwen-7b",
      "developer": "Qwen",
      "benchmark": "MMLU",
      "score": 26.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-28",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen-7B",
      "model_slug": "qwen-7b",
      "developer": "Qwen",
      "benchmark": "PIQA",
      "score": 55.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-28",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Qwen2-72B",
      "model_slug": "qwen2-72b",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 21.92,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-06-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2-72B",
      "model_slug": "qwen2-72b",
      "developer": "Qwen",
      "benchmark": "MATH level 5",
      "score": 39.07,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-06-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2-72B",
      "model_slug": "qwen2-72b",
      "developer": "Qwen",
      "benchmark": "WeirdML",
      "score": 11.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-06-07",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2-72B",
      "model_slug": "qwen2-72b",
      "developer": "Qwen",
      "benchmark": "METR Time Horizons",
      "score": 29.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2024-06-07",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2-72B",
      "model_slug": "qwen2-72b",
      "developer": "Qwen",
      "benchmark": "The Agent Company",
      "score": 1.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "TheAgentCompany experiment results github",
      "measured_at": "2024-06-07",
      "source": "TheAgentCompany experiment results github",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2-72B",
      "model_slug": "qwen2-72b",
      "developer": "Qwen",
      "benchmark": "MMLU",
      "score": 76.53,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-06-07",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "ARC AI2",
      "score": 92.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-09-19",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "Winogrande",
      "score": 64.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-09-19",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 33.54,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "MATH level 5",
      "score": 63.17,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 7.96,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "TriviaQA",
      "score": 71.9,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-09-19",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 38.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-09-19",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "BBH",
      "score": 73.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-09-19",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "WeirdML",
      "score": 15.97,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "WeirdML Leaderboard",
      "measured_at": "2024-09-19",
      "source": "WeirdML Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 15.79,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2024-09-19",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "METR Time Horizons",
      "score": 35.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2024-09-19",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "Balrog",
      "score": 16.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2024-09-19",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "GeoBench",
      "score": 62,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2024-09-19",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "The Agent Company",
      "score": 5.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "TheAgentCompany experiment results github",
      "measured_at": "2024-09-19",
      "source": "TheAgentCompany experiment results github",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "OSWorld",
      "score": 5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "OS World Website",
      "measured_at": "2024-09-19",
      "source": "OS World Website",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "HellaSwag",
      "score": 79.73,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-09-19",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "MMLU",
      "score": 80.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-4 Technical Report",
      "measured_at": "2024-09-19",
      "source": "Phi-4 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-72B",
      "model_slug": "qwen2-5-72b",
      "developer": "Qwen",
      "benchmark": "PIQA",
      "score": 65.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-09-19",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Qwen2.5-Coder (1.5B)",
      "model_slug": "qwen2-5-coder-1-5b",
      "developer": "Qwen",
      "benchmark": "ARC AI2",
      "score": 26.93,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder (1.5B)",
      "model_slug": "qwen2-5-coder-1-5b",
      "developer": "Qwen",
      "benchmark": "Winogrande",
      "score": 21.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder (1.5B)",
      "model_slug": "qwen2-5-coder-1-5b",
      "developer": "Qwen",
      "benchmark": "GSM8K",
      "score": 65.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder (1.5B)",
      "model_slug": "qwen2-5-coder-1-5b",
      "developer": "Qwen",
      "benchmark": "HellaSwag",
      "score": 49.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder (1.5B)",
      "model_slug": "qwen2-5-coder-1-5b",
      "developer": "Qwen",
      "benchmark": "MMLU",
      "score": 38.13,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder (7B)",
      "model_slug": "qwen2-5-coder-7b",
      "developer": "Qwen",
      "benchmark": "ARC AI2",
      "score": 47.87,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder (7B)",
      "model_slug": "qwen2-5-coder-7b",
      "developer": "Qwen",
      "benchmark": "Winogrande",
      "score": 45.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder (7B)",
      "model_slug": "qwen2-5-coder-7b",
      "developer": "Qwen",
      "benchmark": "GSM8K",
      "score": 86.7,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder (7B)",
      "model_slug": "qwen2-5-coder-7b",
      "developer": "Qwen",
      "benchmark": "HellaSwag",
      "score": 69.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder (7B)",
      "model_slug": "qwen2-5-coder-7b",
      "developer": "Qwen",
      "benchmark": "MMLU",
      "score": 57.33,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 37.64,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-24",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "ProofBench",
      "score": 36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2025-11-24",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 84.79,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 86.1,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 83.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-11-24",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 63.74,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-11-24",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 52.31,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-11-24",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 45.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-24",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 9.51,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-24",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "Cybench",
      "score": 82,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Opus 4.5 System Card",
      "measured_at": "2025-11-24",
      "source": "Opus 4.5 System Card",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 54.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-11-24",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "METR Time Horizons",
      "score": 74.97,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-11-24",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "Balrog",
      "score": 43.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Balrog Leaderboard",
      "measured_at": "2025-11-24",
      "source": "Balrog Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "GeoBench",
      "score": 75,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2025-11-24",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "VPCT",
      "score": 10,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Tweet from VPCT creator",
      "measured_at": "2025-11-24",
      "source": "Tweet from VPCT creator",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "GSO-Bench",
      "score": 26.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GSO Leaderboard",
      "measured_at": "2025-11-24",
      "source": "GSO Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "DeepResearch Bench",
      "score": 54.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2025-11-24",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "OSWorld",
      "score": 66.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Anthropic annoucement",
      "measured_at": "2025-11-24",
      "source": "Anthropic annoucement",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 36.2,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 4.88,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "EBR-bench",
      "score": 14.29,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "Mystery Game Puzzles",
      "score": 14.08,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "Remote Labor Index",
      "score": 3.75,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-24",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "Furniture Assembly",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-24",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "WeirdML v3",
      "score": 7.92,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-11-24",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "Terminal Bench",
      "score": 63.15,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2025-11-24",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "GDPval",
      "score": 45.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-24",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "CL-bench",
      "score": 21.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-24",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 80,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-11-24",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.5",
      "model_slug": "claude-opus-4-5",
      "developer": "Anthropic",
      "benchmark": "SWE-Bench verified",
      "score": 76.65,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-0.5B",
      "model_slug": "qwen2-5-coder-0-5b",
      "developer": "Qwen",
      "benchmark": "ARC AI2",
      "score": 12.53,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-0.5B",
      "model_slug": "qwen2-5-coder-0-5b",
      "developer": "Qwen",
      "benchmark": "Winogrande",
      "score": 9.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-0.5B",
      "model_slug": "qwen2-5-coder-0-5b",
      "developer": "Qwen",
      "benchmark": "GSM8K",
      "score": 34.5,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-0.5B",
      "model_slug": "qwen2-5-coder-0-5b",
      "developer": "Qwen",
      "benchmark": "HellaSwag",
      "score": 31.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-0.5B",
      "model_slug": "qwen2-5-coder-0-5b",
      "developer": "Qwen",
      "benchmark": "MMLU",
      "score": 22.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-14B",
      "model_slug": "qwen2-5-coder-14b",
      "developer": "Qwen",
      "benchmark": "ARC AI2",
      "score": 54.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-14B",
      "model_slug": "qwen2-5-coder-14b",
      "developer": "Qwen",
      "benchmark": "Winogrande",
      "score": 53.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-14B",
      "model_slug": "qwen2-5-coder-14b",
      "developer": "Qwen",
      "benchmark": "GSM8K",
      "score": 88.7,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-14B",
      "model_slug": "qwen2-5-coder-14b",
      "developer": "Qwen",
      "benchmark": "HellaSwag",
      "score": 73.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-14B",
      "model_slug": "qwen2-5-coder-14b",
      "developer": "Qwen",
      "benchmark": "MMLU",
      "score": 66.93,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-32B",
      "model_slug": "qwen2-5-coder-32b",
      "developer": "Qwen",
      "benchmark": "ARC AI2",
      "score": 60.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-32B",
      "model_slug": "qwen2-5-coder-32b",
      "developer": "Qwen",
      "benchmark": "Winogrande",
      "score": 61.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-32B",
      "model_slug": "qwen2-5-coder-32b",
      "developer": "Qwen",
      "benchmark": "GSM8K",
      "score": 91.1,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-32B",
      "model_slug": "qwen2-5-coder-32b",
      "developer": "Qwen",
      "benchmark": "HellaSwag",
      "score": 77.33,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-32B",
      "model_slug": "qwen2-5-coder-32b",
      "developer": "Qwen",
      "benchmark": "MMLU",
      "score": 72.13,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-3B",
      "model_slug": "qwen2-5-coder-3b",
      "developer": "Qwen",
      "benchmark": "ARC AI2",
      "score": 37.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-3B",
      "model_slug": "qwen2-5-coder-3b",
      "developer": "Qwen",
      "benchmark": "Winogrande",
      "score": 34.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-3B",
      "model_slug": "qwen2-5-coder-3b",
      "developer": "Qwen",
      "benchmark": "GSM8K",
      "score": 75.7,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Coder-3B",
      "model_slug": "qwen2-5-coder-3b",
      "developer": "Qwen",
      "benchmark": "HellaSwag",
      "score": 61.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-09-18",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen2.5-Max",
      "model_slug": "qwen2-5-max",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 43.23,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-01-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2.5-Max",
      "model_slug": "qwen2-5-max",
      "developer": "Qwen",
      "benchmark": "MATH level 5",
      "score": 67.18,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-01-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2.5-Max",
      "model_slug": "qwen2-5-max",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 16.03,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-01-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2.5-Max",
      "model_slug": "qwen2-5-max",
      "developer": "Qwen",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 1.81,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-01-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen2.5-Max",
      "model_slug": "qwen2-5-max",
      "developer": "Qwen",
      "benchmark": "Aider polyglot",
      "score": 21.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-01-25",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2.5-Max",
      "model_slug": "qwen2-5-max",
      "developer": "Qwen",
      "benchmark": "Fiction.LiveBench",
      "score": 66.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-01-25",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen2.5-Max",
      "model_slug": "qwen2-5-max",
      "developer": "Qwen",
      "benchmark": "Lech Mazur Writing",
      "score": 72.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-01-25",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen3-14B",
      "model_slug": "qwen3-14b",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 53.84,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-14B",
      "model_slug": "qwen3-14b",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 66.36,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-14B",
      "model_slug": "qwen3-14b",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 40,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-04-29",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-14B",
      "model_slug": "qwen3-14b",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 21.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-04-29",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-14B",
      "model_slug": "qwen3-14b",
      "developer": "Qwen",
      "benchmark": "Fiction.LiveBench",
      "score": 62.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-04-28",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-235B-A22B-Instruct (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-instruct-jul-2025",
      "developer": "Qwen",
      "benchmark": "ARC-AGI-2",
      "score": 1.25,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-07-25",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-235B-A22B-Instruct (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-instruct-jul-2025",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 64,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-07-25",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-235B-A22B-Instruct (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-instruct-jul-2025",
      "developer": "Qwen",
      "benchmark": "WeirdML",
      "score": 38.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "WeirdML Leaderboard",
      "measured_at": "2025-07-25",
      "source": "WeirdML Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-235B-A22B-Instruct (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-instruct-jul-2025",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 27.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-07-25",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-235B-A22B-Instruct (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-instruct-jul-2025",
      "developer": "Qwen",
      "benchmark": "Aider polyglot",
      "score": 59.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-07-25",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-235B-A22B-Instruct (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-instruct-jul-2025",
      "developer": "Qwen",
      "benchmark": "Fiction.LiveBench",
      "score": 52.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-07-25",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-235B-A22B-Instruct (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-instruct-jul-2025",
      "developer": "Qwen",
      "benchmark": "ARC-AGI",
      "score": 11,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-07-25",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen3-235B-A22B-Thinking (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-thinking-jul-2025",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 76.46,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-07-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-235B-A22B-Thinking (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-thinking-jul-2025",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 86.65,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-07-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-235B-A22B-Thinking (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-thinking-jul-2025",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 67.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-07-25",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-235B-A22B-Thinking (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-thinking-jul-2025",
      "developer": "Qwen",
      "benchmark": "WeirdML",
      "score": 41.04,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "WeirdML Leaderboard",
      "measured_at": "2025-07-25",
      "source": "WeirdML Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-235B-A22B-Thinking (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-thinking-jul-2025",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 34.48,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-07-25",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-235B-A22B-Thinking (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-thinking-jul-2025",
      "developer": "Qwen",
      "benchmark": "SimpleQA Verified",
      "score": 40.44,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-07-25",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-235B-A22B-Thinking (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-thinking-jul-2025",
      "developer": "Qwen",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 14.88,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-07-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-235B-A22B-Thinking (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-thinking-jul-2025",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-07-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-235B-A22B-Thinking (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-thinking-jul-2025",
      "developer": "Qwen",
      "benchmark": "Fiction.LiveBench",
      "score": 75,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-07-25",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-235B-A22B-Thinking (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-thinking-jul-2025",
      "developer": "Qwen",
      "benchmark": "Mystery Game Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-07-25",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-235B-A22B-Thinking (Jul 2025)",
      "model_slug": "qwen3-235b-a22b-thinking-jul-2025",
      "developer": "Qwen",
      "benchmark": "Lech Mazur Writing",
      "score": 82.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-07-25",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen3-30B-A3B",
      "model_slug": "qwen3-30b-a3b",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 50.94,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-30B-A3B",
      "model_slug": "qwen3-30b-a3b",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 62.74,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-30B-A3B",
      "model_slug": "qwen3-30b-a3b",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 33.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-04-28",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-30B-A3B",
      "model_slug": "qwen3-30b-a3b",
      "developer": "Qwen",
      "benchmark": "WeirdML",
      "score": 29.75,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "WeirdML Leaderboard",
      "measured_at": "2025-04-28",
      "source": "WeirdML Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-30B-A3B",
      "model_slug": "qwen3-30b-a3b",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 18.61,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-04-28",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-30B-A3B",
      "model_slug": "qwen3-30b-a3b",
      "developer": "Qwen",
      "benchmark": "Fiction.LiveBench",
      "score": 40.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-04-28",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-30B-A3B",
      "model_slug": "qwen3-30b-a3b",
      "developer": "Qwen",
      "benchmark": "Lech Mazur Writing",
      "score": 75.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-04-28",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 69.17,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "ProofBench",
      "score": 50,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-02-05",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 91.01,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 94.44,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 85.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-02-05",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 77.95,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-02-05",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 65.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-02-05",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 47,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 16.88,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "Cybench",
      "score": 93,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Opus 4.6 System Card",
      "measured_at": "2026-02-05",
      "source": "Opus 4.6 System Card",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 61.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-02-05",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "METR Time Horizons",
      "score": 78.86,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-05",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "GSO-Bench",
      "score": 41.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GSO Leaderboard",
      "measured_at": "2026-02-05",
      "source": "GSO Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "DeepResearch Bench",
      "score": 55.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2026-02-05",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 69.44,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 26.83,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "EBR-bench",
      "score": 12.7,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "Mystery Game Puzzles",
      "score": 17.38,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "MirrorCode",
      "score": 33.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "Remote Labor Index",
      "score": 4.17,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "FrontierCode",
      "score": 26.64,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-02-05",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "Furniture Assembly",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "Terminal Bench",
      "score": 79.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2026-02-05",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "CL-bench",
      "score": 20.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-05",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "CL-bench Life",
      "score": 17,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 94,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-02-05",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "APEX-Agents",
      "score": 46.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.6",
      "model_slug": "claude-opus-4-6",
      "developer": "Anthropic",
      "benchmark": "SWE-Bench verified",
      "score": 78.72,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen3-30B-A3B-Instruct (Jul 2025)",
      "model_slug": "qwen3-30b-a3b-instruct-jul-2025",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 42.53,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-07-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-30B-A3B-Instruct (Jul 2025)",
      "model_slug": "qwen3-30b-a3b-instruct-jul-2025",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 62.18,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-07-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-30B-A3B-Instruct (Jul 2025)",
      "model_slug": "qwen3-30b-a3b-instruct-jul-2025",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 45.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-07-29",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-30B-A3B-Instruct (Jul 2025)",
      "model_slug": "qwen3-30b-a3b-instruct-jul-2025",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 26.32,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-07-29",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-30B-A3B-Thinking (Jul 2025)",
      "model_slug": "qwen3-30b-a3b-thinking-jul-2025",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 62.61,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-07-30",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-30B-A3B-Thinking (Jul 2025)",
      "model_slug": "qwen3-30b-a3b-thinking-jul-2025",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 70.25,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-07-30",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-30B-A3B-Thinking (Jul 2025)",
      "model_slug": "qwen3-30b-a3b-thinking-jul-2025",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 48.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-07-30",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-30B-A3B-Thinking (Jul 2025)",
      "model_slug": "qwen3-30b-a3b-thinking-jul-2025",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 22.98,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-07-30",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-32B",
      "model_slug": "qwen3-32b",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 56.56,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-32B",
      "model_slug": "qwen3-32b",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 66.91,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-32B",
      "model_slug": "qwen3-32b",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 45.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-04-28",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-32B",
      "model_slug": "qwen3-32b",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 20.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-04-28",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-32B",
      "model_slug": "qwen3-32b",
      "developer": "Qwen",
      "benchmark": "Aider polyglot",
      "score": 40,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-04-28",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-32B",
      "model_slug": "qwen3-32b",
      "developer": "Qwen",
      "benchmark": "Fiction.LiveBench",
      "score": 74.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-04-28",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-8B",
      "model_slug": "qwen3-8b",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 44.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-28",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-8B",
      "model_slug": "qwen3-8b",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 56.07,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-28",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-8B",
      "model_slug": "qwen3-8b",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 32.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-04-28",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-8B",
      "model_slug": "qwen3-8b",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 10.38,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-04-28",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-8B",
      "model_slug": "qwen3-8b",
      "developer": "Qwen",
      "benchmark": "Fiction.LiveBench",
      "score": 62.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-04-28",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-Max",
      "model_slug": "qwen3-max",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 66.11,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-09-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-Max",
      "model_slug": "qwen3-max",
      "developer": "Qwen",
      "benchmark": "MATH level 5",
      "score": 97.13,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-09-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-Max",
      "model_slug": "qwen3-max",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 73.31,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-09-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-Max",
      "model_slug": "qwen3-max",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 70.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-09-24",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-Max",
      "model_slug": "qwen3-max",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 33.26,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-09-24",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-Max",
      "model_slug": "qwen3-max",
      "developer": "Qwen",
      "benchmark": "SimpleQA Verified",
      "score": 48.75,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-09-24",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-Max",
      "model_slug": "qwen3-max",
      "developer": "Qwen",
      "benchmark": "Fiction.LiveBench",
      "score": 66.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-09-24",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3-Max",
      "model_slug": "qwen3-max",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 19.94,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-09-24",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-Max",
      "model_slug": "qwen3-max",
      "developer": "Qwen",
      "benchmark": "Mystery Game Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-09-24",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3-Max",
      "model_slug": "qwen3-max",
      "developer": "Qwen",
      "benchmark": "CL-bench",
      "score": 14.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-09-24",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen3.5 397B-A17B",
      "model_slug": "qwen3-5-397b-a17b",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 85.23,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-13",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.5 397B-A17B",
      "model_slug": "qwen3-5-397b-a17b",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 88.88,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-13",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.5 397B-A17B",
      "model_slug": "qwen3-5-397b-a17b",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 79.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-02-13",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.5 397B-A17B",
      "model_slug": "qwen3-5-397b-a17b",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 44.58,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-02-13",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.5 397B-A17B",
      "model_slug": "qwen3-5-397b-a17b",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 32.87,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-13",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.5 397B-A17B",
      "model_slug": "qwen3-5-397b-a17b",
      "developer": "Qwen",
      "benchmark": "Mystery Game Puzzles",
      "score": 9.67,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-13",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.5 397B-A17B",
      "model_slug": "qwen3-5-397b-a17b",
      "developer": "Qwen",
      "benchmark": "APEX-Agents",
      "score": 24.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-13",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.5-35B-A3B",
      "model_slug": "qwen3-5-35b-a3b",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 81.19,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-24",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.5-35B-A3B",
      "model_slug": "qwen3-5-35b-a3b",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 69.97,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-24",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.5-35B-A3B",
      "model_slug": "qwen3-5-35b-a3b",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 66.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-02-24",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.5-35B-A3B",
      "model_slug": "qwen3-5-35b-a3b",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 34.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-02-24",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.5-9B",
      "model_slug": "qwen3-5-9b",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 74.97,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-24",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.5-9B",
      "model_slug": "qwen3-5-9b",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 61.63,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-24",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.5-9B",
      "model_slug": "qwen3-5-9b",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 52,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-02-24",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.5-9B",
      "model_slug": "qwen3-5-9b",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 28.79,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-02-24",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.5-9B",
      "model_slug": "qwen3-5-9b",
      "developer": "Qwen",
      "benchmark": "Terminal Bench",
      "score": 9.21,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2026-02-24",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.6 27B",
      "model_slug": "qwen3-6-27b",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 84.53,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.6 27B",
      "model_slug": "qwen3-6-27b",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 91.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.6 27B",
      "model_slug": "qwen3-6-27b",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 63.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-04-22",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.6 27B",
      "model_slug": "qwen3-6-27b",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 40.62,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-04-22",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.6 27B",
      "model_slug": "qwen3-6-27b",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 36.93,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.6 27B",
      "model_slug": "qwen3-6-27b",
      "developer": "Qwen",
      "benchmark": "Mystery Game Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.7-Max",
      "model_slug": "qwen3-7-max",
      "developer": "Qwen",
      "benchmark": "ProofBench",
      "score": 26,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-05-19",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.7-Max",
      "model_slug": "qwen3-7-max",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 91.54,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.7-Max",
      "model_slug": "qwen3-7-max",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 95.55,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.7-Max",
      "model_slug": "qwen3-7-max",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 87.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-05-19",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.7-Max",
      "model_slug": "qwen3-7-max",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 51.79,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-05-19",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.7-Max",
      "model_slug": "qwen3-7-max",
      "developer": "Qwen",
      "benchmark": "SimpleQA Verified",
      "score": 55.77,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-05-19",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.7-Max",
      "model_slug": "qwen3-7-max",
      "developer": "Qwen",
      "benchmark": "Chess Puzzles",
      "score": 23.19,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-05-19",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.7-Max",
      "model_slug": "qwen3-7-max",
      "developer": "Qwen",
      "benchmark": "SimpleBench",
      "score": 64.48,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2026-05-19",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.7-Max",
      "model_slug": "qwen3-7-max",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 67.96,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.7-Max",
      "model_slug": "qwen3-7-max",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 34.15,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.7-Max",
      "model_slug": "qwen3-7-max",
      "developer": "Qwen",
      "benchmark": "EBR-bench",
      "score": 9.52,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.7-Max",
      "model_slug": "qwen3-7-max",
      "developer": "Qwen",
      "benchmark": "Mystery Game Puzzles",
      "score": 25.09,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-05-19",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.7-Max",
      "model_slug": "qwen3-7-max",
      "developer": "Qwen",
      "benchmark": "SWE-Bench verified",
      "score": 77.27,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-19",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 75.83,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "ProofBench",
      "score": 54,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-04-16",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 90.49,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 97.8,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 91.08,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-04-16",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 76.44,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-04-16",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 61.41,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-04-16",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 51.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 37.92,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 54.04,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-04-16",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "GSO-Bench",
      "score": 44.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2026-04-16",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 73.87,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 31.71,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "EBR-bench",
      "score": 19.05,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "Mystery Game Puzzles",
      "score": 20.69,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "MirrorCode",
      "score": 31.11,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "PostTrainBench",
      "score": 24.34,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "FrontierCode",
      "score": 38.54,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-04-16",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "Furniture Assembly",
      "score": 4.76,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "Terminal Bench",
      "score": 80.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2026-04-16",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "ExploitBench",
      "score": 26.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "OSWorld 2.0",
      "score": 18.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://osworld-v2.xlang.ai/",
      "measured_at": "2026-04-16",
      "source": "https://osworld-v2.xlang.ai/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 93.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-04-16",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "APEX-Agents",
      "score": 49.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.7",
      "model_slug": "claude-opus-4-7",
      "developer": "Anthropic",
      "benchmark": "SWE-Bench verified",
      "score": 83.47,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Qwen3.7-Plus",
      "model_slug": "qwen3-7-plus",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 87.33,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-02",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.7-Plus",
      "model_slug": "qwen3-7-plus",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 93.33,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-02",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.7-Plus",
      "model_slug": "qwen3-7-plus",
      "developer": "Qwen",
      "benchmark": "DTBench",
      "score": 73.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-06-02",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.7-Plus",
      "model_slug": "qwen3-7-plus",
      "developer": "Qwen",
      "benchmark": "LMCA",
      "score": 44.21,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-06-02",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.7-Plus",
      "model_slug": "qwen3-7-plus",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 36.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.7-Plus",
      "model_slug": "qwen3-7-plus",
      "developer": "Qwen",
      "benchmark": "Mystery Game Puzzles",
      "score": 8.57,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-02",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.7-Plus",
      "model_slug": "qwen3-7-plus",
      "developer": "Qwen",
      "benchmark": "FrontierCode",
      "score": 10.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-06-02",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.7-Plus",
      "model_slug": "qwen3-7-plus",
      "developer": "Qwen",
      "benchmark": "OSWorld 2.0",
      "score": 2.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://osworld-v2.xlang.ai/",
      "measured_at": "2026-06-02",
      "source": "https://osworld-v2.xlang.ai/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.8 Max (0902)",
      "model_slug": "qwen3-8-max-0902",
      "developer": "Qwen",
      "benchmark": "GPQA diamond",
      "score": 93.47,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.8 Max (0902)",
      "model_slug": "qwen3-8-max-0902",
      "developer": "Qwen",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.8 Max (0902)",
      "model_slug": "qwen3-8-max-0902",
      "developer": "Qwen",
      "benchmark": "SimpleQA Verified",
      "score": 47.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Qwen3.8 Max (0902)",
      "model_slug": "qwen3-8-max-0902",
      "developer": "Qwen",
      "benchmark": "Chess Puzzles",
      "score": 33.71,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.8 Max (0902)",
      "model_slug": "qwen3-8-max-0902",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 69.07,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.8 Max (0902)",
      "model_slug": "qwen3-8-max-0902",
      "developer": "Qwen",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 34.15,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.8 Max (0902)",
      "model_slug": "qwen3-8-max-0902",
      "developer": "Qwen",
      "benchmark": "MirrorCode",
      "score": 47.78,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.8 Max (0902)",
      "model_slug": "qwen3-8-max-0902",
      "developer": "Qwen",
      "benchmark": "Furniture Assembly",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-01",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Qwen3.8 Max (0902)",
      "model_slug": "qwen3-8-max-0902",
      "developer": "Qwen",
      "benchmark": "FrontierSWE",
      "score": 17.81,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-09-01",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "RedPajama-INCITE-7B-Base",
      "model_slug": "redpajama-incite-7b-base",
      "developer": "Together AI",
      "benchmark": "ARC AI2",
      "score": 18.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-05-04",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "RedPajama-INCITE-7B-Base",
      "model_slug": "redpajama-incite-7b-base",
      "developer": "Together AI",
      "benchmark": "OpenBookQA",
      "score": 20,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-05-04",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "RedPajama-INCITE-7B-Base",
      "model_slug": "redpajama-incite-7b-base",
      "developer": "Together AI",
      "benchmark": "Winogrande",
      "score": 27.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-05-04",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "RedPajama-INCITE-7B-Base",
      "model_slug": "redpajama-incite-7b-base",
      "developer": "Together AI",
      "benchmark": "HellaSwag",
      "score": 60.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-05-04",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "RedPajama-INCITE-7B-Base",
      "model_slug": "redpajama-incite-7b-base",
      "developer": "Together AI",
      "benchmark": "MMLU",
      "score": 1.73,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-05-04",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "RedPajama-INCITE-7B-Base",
      "model_slug": "redpajama-incite-7b-base",
      "developer": "Together AI",
      "benchmark": "PIQA",
      "score": 53.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-05-04",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Stable Beluga 2",
      "model_slug": "stable-beluga-2",
      "developer": "Stability AI",
      "benchmark": "ARC AI2",
      "score": 81.47,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-20",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Stable Beluga 2",
      "model_slug": "stable-beluga-2",
      "developer": "Stability AI",
      "benchmark": "BBH",
      "score": 59.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-20",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Stable Beluga 2",
      "model_slug": "stable-beluga-2",
      "developer": "Stability AI",
      "benchmark": "GSM8K",
      "score": 69.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-20",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Stable Beluga 2",
      "model_slug": "stable-beluga-2",
      "developer": "Stability AI",
      "benchmark": "HellaSwag",
      "score": 78.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-20",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Stable Beluga 2",
      "model_slug": "stable-beluga-2",
      "developer": "Stability AI",
      "benchmark": "LAMBADA",
      "score": 71.3,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-20",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Stable Beluga 2",
      "model_slug": "stable-beluga-2",
      "developer": "Stability AI",
      "benchmark": "MMLU",
      "score": 58.13,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-20",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Stable Beluga 2",
      "model_slug": "stable-beluga-2",
      "developer": "Stability AI",
      "benchmark": "PIQA",
      "score": 66.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-20",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "StarCoder 2 15B",
      "model_slug": "starcoder-2-15b",
      "developer": "BigCode",
      "benchmark": "ARC AI2",
      "score": 29.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-02-20",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "StarCoder 2 15B",
      "model_slug": "starcoder-2-15b",
      "developer": "BigCode",
      "benchmark": "Winogrande",
      "score": 28.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-02-20",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "StarCoder 2 15B",
      "model_slug": "starcoder-2-15b",
      "developer": "BigCode",
      "benchmark": "GSM8K",
      "score": 57.7,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-02-20",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "StarCoder 2 15B",
      "model_slug": "starcoder-2-15b",
      "developer": "BigCode",
      "benchmark": "MMLU",
      "score": 52.13,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-02-20",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "StarCoder 2 3B",
      "model_slug": "starcoder-2-3b",
      "developer": "BigCode",
      "benchmark": "ARC AI2",
      "score": 12.27,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-02-22",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "StarCoder 2 3B",
      "model_slug": "starcoder-2-3b",
      "developer": "BigCode",
      "benchmark": "Winogrande",
      "score": 14.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-02-22",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "StarCoder 2 3B",
      "model_slug": "starcoder-2-3b",
      "developer": "BigCode",
      "benchmark": "GSM8K",
      "score": 21.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-02-22",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "StarCoder 2 3B",
      "model_slug": "starcoder-2-3b",
      "developer": "BigCode",
      "benchmark": "MMLU",
      "score": 15.47,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-02-22",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "StarCoder 2 7B",
      "model_slug": "starcoder-2-7b",
      "developer": "BigCode",
      "benchmark": "ARC AI2",
      "score": 18.27,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-02-20",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "StarCoder 2 7B",
      "model_slug": "starcoder-2-7b",
      "developer": "BigCode",
      "benchmark": "Winogrande",
      "score": 14.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-02-20",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "StarCoder 2 7B",
      "model_slug": "starcoder-2-7b",
      "developer": "BigCode",
      "benchmark": "GSM8K",
      "score": 32.7,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-02-20",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "StarCoder 2 7B",
      "model_slug": "starcoder-2-7b",
      "developer": "BigCode",
      "benchmark": "MMLU",
      "score": 18.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-02-20",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "XGen-7B",
      "model_slug": "xgen-7b",
      "developer": "Salesforce AI Research",
      "benchmark": "ARC AI2",
      "score": 21.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-06-27",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "XGen-7B",
      "model_slug": "xgen-7b",
      "developer": "Salesforce AI Research",
      "benchmark": "OpenBookQA",
      "score": 20.27,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-06-27",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "XGen-7B",
      "model_slug": "xgen-7b",
      "developer": "Salesforce AI Research",
      "benchmark": "Winogrande",
      "score": 29.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-06-27",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "XGen-7B",
      "model_slug": "xgen-7b",
      "developer": "Salesforce AI Research",
      "benchmark": "HellaSwag",
      "score": 65.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-06-27",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "XGen-7B",
      "model_slug": "xgen-7b",
      "developer": "Salesforce AI Research",
      "benchmark": "MMLU",
      "score": 15.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-06-27",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "XGen-7B",
      "model_slug": "xgen-7b",
      "developer": "Salesforce AI Research",
      "benchmark": "PIQA",
      "score": 51,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-06-27",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Yi 6B",
      "model_slug": "yi-6b",
      "developer": "01.AI",
      "benchmark": "ARC AI2",
      "score": 33.73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2023-11-02",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Yi 6B",
      "model_slug": "yi-6b",
      "developer": "01.AI",
      "benchmark": "Winogrande",
      "score": 42.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2023-11-02",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Yi 6B",
      "model_slug": "yi-6b",
      "developer": "01.AI",
      "benchmark": "BBH",
      "score": 29.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2023-11-22",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Yi 6B",
      "model_slug": "yi-6b",
      "developer": "01.AI",
      "benchmark": "GSM8K",
      "score": 44.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2023-11-22",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Yi 6B",
      "model_slug": "yi-6b",
      "developer": "01.AI",
      "benchmark": "HellaSwag",
      "score": 65.87,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2023-11-02",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Yi 6B",
      "model_slug": "yi-6b",
      "developer": "01.AI",
      "benchmark": "MMLU",
      "score": 52,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2023-11-22",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Yi-9B",
      "model_slug": "yi-9b",
      "developer": "01.AI",
      "benchmark": "ARC AI2",
      "score": 40.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2024-03-01",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Yi-9B",
      "model_slug": "yi-9b",
      "developer": "01.AI",
      "benchmark": "Winogrande",
      "score": 46,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2024-03-01",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Yi-9B",
      "model_slug": "yi-9b",
      "developer": "01.AI",
      "benchmark": "HellaSwag",
      "score": 68.53,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2024-03-01",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Yi-9B",
      "model_slug": "yi-9b",
      "developer": "01.AI",
      "benchmark": "MMLU",
      "score": 57.87,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2024-03-01",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 72.08,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-05-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "ProofBench",
      "score": 69,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-05-28",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 91.72,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-28",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 98.33,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-28",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 91.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-05-28",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 82.89,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-05-28",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 67.66,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-05-28",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 53,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-05-28",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 49.49,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-05-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 57.76,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-05-28",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "GeoBench",
      "score": 81,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2026-05-28",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "GSO-Bench",
      "score": 47.06,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2026-05-28",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "DeepResearch Bench",
      "score": 50.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2026-05-28",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 84.21,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-28",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 56.1,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-28",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "EBR-bench",
      "score": 28.57,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-28",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "Mystery Game Puzzles",
      "score": 29.5,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-05-28",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "MirrorCode",
      "score": 65.56,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-05-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "PostTrainBench",
      "score": 31,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-05-28",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "Remote Labor Index",
      "score": 8.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-05-28",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "Surface Evolver Bench",
      "score": 87.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-05-28",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "FrontierCode",
      "score": 46.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-05-28",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "Furniture Assembly",
      "score": 17.86,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-05-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "OSWorld 2.0",
      "score": 20.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://osworld-v2.xlang.ai/",
      "measured_at": "2026-05-28",
      "source": "https://osworld-v2.xlang.ai/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 92.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-05-28",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 4.8",
      "model_slug": "claude-opus-4-8",
      "developer": "Anthropic",
      "benchmark": "APEX-Agents",
      "score": 48.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-05-28",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "chatglm2-6b",
      "model_slug": "chatglm2-6b",
      "developer": "Zhipu AI",
      "benchmark": "ARC AI2",
      "score": 48,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-06-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "chatglm2-6b",
      "model_slug": "chatglm2-6b",
      "developer": "Zhipu AI",
      "benchmark": "BBH",
      "score": 11.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-06-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "chatglm2-6b",
      "model_slug": "chatglm2-6b",
      "developer": "Zhipu AI",
      "benchmark": "GSM8K",
      "score": 32.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-06-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "chatglm2-6b",
      "model_slug": "chatglm2-6b",
      "developer": "Zhipu AI",
      "benchmark": "HellaSwag",
      "score": 42.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-06-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "chatglm2-6b",
      "model_slug": "chatglm2-6b",
      "developer": "Zhipu AI",
      "benchmark": "LAMBADA",
      "score": 54.3,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-06-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "chatglm2-6b",
      "model_slug": "chatglm2-6b",
      "developer": "Zhipu AI",
      "benchmark": "MMLU",
      "score": 30.53,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-06-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "chatglm2-6b",
      "model_slug": "chatglm2-6b",
      "developer": "Zhipu AI",
      "benchmark": "PIQA",
      "score": 39.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-06-24",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "internlm-20b",
      "model_slug": "internlm-20b",
      "developer": "Shanghai AI Laboratory",
      "benchmark": "ARC AI2",
      "score": 75.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-18",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "internlm-20b",
      "model_slug": "internlm-20b",
      "developer": "Shanghai AI Laboratory",
      "benchmark": "BBH",
      "score": 36.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-18",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "internlm-20b",
      "model_slug": "internlm-20b",
      "developer": "Shanghai AI Laboratory",
      "benchmark": "GSM8K",
      "score": 62.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2023-09-18",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "internlm-20b",
      "model_slug": "internlm-20b",
      "developer": "Shanghai AI Laboratory",
      "benchmark": "HellaSwag",
      "score": 70.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-18",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "internlm-20b",
      "model_slug": "internlm-20b",
      "developer": "Shanghai AI Laboratory",
      "benchmark": "LAMBADA",
      "score": 71.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-18",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "internlm-20b",
      "model_slug": "internlm-20b",
      "developer": "Shanghai AI Laboratory",
      "benchmark": "PIQA",
      "score": 60.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-09-18",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "internlm-7b",
      "model_slug": "internlm-7b",
      "developer": "Shanghai AI Laboratory",
      "benchmark": "ARC AI2",
      "score": 59.33,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-05",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "internlm-7b",
      "model_slug": "internlm-7b",
      "developer": "Shanghai AI Laboratory",
      "benchmark": "BBH",
      "score": 16,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-05",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "internlm-7b",
      "model_slug": "internlm-7b",
      "developer": "Shanghai AI Laboratory",
      "benchmark": "GSM8K",
      "score": 31.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-05",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "internlm-7b",
      "model_slug": "internlm-7b",
      "developer": "Shanghai AI Laboratory",
      "benchmark": "HellaSwag",
      "score": 60.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-05",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "internlm-7b",
      "model_slug": "internlm-7b",
      "developer": "Shanghai AI Laboratory",
      "benchmark": "LAMBADA",
      "score": 67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-05",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "internlm-7b",
      "model_slug": "internlm-7b",
      "developer": "Shanghai AI Laboratory",
      "benchmark": "MMLU",
      "score": 34.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-05",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "internlm-7b",
      "model_slug": "internlm-7b",
      "developer": "Shanghai AI Laboratory",
      "benchmark": "PIQA",
      "score": 55.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-07-05",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 71.9,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 94.71,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 73.31,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-12-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 57.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-12-17",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "CadEval",
      "score": 56,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "CadEval Dashboard",
      "measured_at": "2024-12-17",
      "source": "CadEval Dashboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 46.08,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-12-17",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 26.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2024-12-17",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 41.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-12-17",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 6.36,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-12-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 28.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2024-12-17",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "METR Time Horizons",
      "score": 51.05,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "measured_at": "2024-12-17",
      "source": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "Aider polyglot",
      "score": 61.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2024-12-17",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "GeoBench",
      "score": 80,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2024-12-17",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "VPCT",
      "score": 5.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2024-12-17",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "Fiction.LiveBench",
      "score": 83.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2024-12-17",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 15.51,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-12-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 30.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2024-12-17",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o1",
      "model_slug": "o1",
      "developer": "OpenAI",
      "benchmark": "Lech Mazur Writing",
      "score": 70.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2024-12-17",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o1-mini",
      "model_slug": "o1-mini",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 0.83,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-09-12",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o1-mini",
      "model_slug": "o1-mini",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 51.91,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1-mini",
      "model_slug": "o1-mini",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 89.18,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1-mini",
      "model_slug": "o1-mini",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 46.89,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1-mini",
      "model_slug": "o1-mini",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 36.32,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-09-12",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1-mini",
      "model_slug": "o1-mini",
      "developer": "OpenAI",
      "benchmark": "Cybench",
      "score": 10,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Cybench leaderboard",
      "measured_at": "2024-09-12",
      "source": "Cybench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1-mini",
      "model_slug": "o1-mini",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 1.72,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2024-09-12",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o1-mini",
      "model_slug": "o1-mini",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 3.02,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o1-mini",
      "model_slug": "o1-mini",
      "developer": "OpenAI",
      "benchmark": "Aider polyglot",
      "score": 32.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2024-09-12",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1-mini",
      "model_slug": "o1-mini",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 14,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2024-09-12",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o1-mini",
      "model_slug": "o1-mini",
      "developer": "OpenAI",
      "benchmark": "Lech Mazur Writing",
      "score": 64.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2024-09-12",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o1-preview",
      "model_slug": "o1-preview",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 35.16,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1-preview",
      "model_slug": "o1-preview",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 81.65,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1-preview",
      "model_slug": "o1-preview",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 31.04,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-12",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1-preview",
      "model_slug": "o1-preview",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 47.56,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-09-12",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1-preview",
      "model_slug": "o1-preview",
      "developer": "OpenAI",
      "benchmark": "Cybench",
      "score": 10,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Cybench leaderboard",
      "measured_at": "2024-09-12",
      "source": "Cybench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o1-preview",
      "model_slug": "o1-preview",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 30.04,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2024-09-12",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o1-preview",
      "model_slug": "o1-preview",
      "developer": "OpenAI",
      "benchmark": "METR Time Horizons",
      "score": 45.14,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "measured_at": "2024-09-12",
      "source": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o1-preview",
      "model_slug": "o1-preview",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 18,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2024-09-12",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 6.53,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 78.91,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 97.77,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 84.43,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 74.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-04-16",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "CadEval",
      "score": 74,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "CadEval Dashboard",
      "measured_at": "2025-04-16",
      "source": "CadEval Dashboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 52.42,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-04-16",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 46.71,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-04-16",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 49.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 33.71,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 43.72,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-04-16",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "METR Time Horizons",
      "score": 65.44,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-04-16",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "Aider polyglot",
      "score": 81.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-04-16",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "GeoBench",
      "score": 74,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2025-04-16",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 3.47,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "VPCT",
      "score": 28,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-04-16",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "GSO-Bench",
      "score": 8.82,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2025-04-16",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "DeepResearch Bench",
      "score": 45.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2025-04-16",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "Fiction.LiveBench",
      "score": 88.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-04-16",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "OSWorld",
      "score": 23,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "OS World Website",
      "measured_at": "2025-04-16",
      "source": "OS World Website",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 35.09,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 21.79,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "GDPval",
      "score": 30.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "CL-bench",
      "score": 17.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 60.83,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2025-04-16",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "Lech Mazur Writing",
      "score": 83.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-04-16",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o3",
      "model_slug": "o3",
      "developer": "OpenAI",
      "benchmark": "SWE-Bench verified",
      "score": 62.32,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 2.99,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-01-31",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 72.25,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-01-31",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 96.49,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-01-31",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 76.92,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-01-31",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 48,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-01-31",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "CadEval",
      "score": 54,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "CadEval Dashboard",
      "measured_at": "2025-01-31",
      "source": "CadEval Dashboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 43.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-01-31",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 22.32,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-01-31",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 15.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-01-31",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 10.56,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-01-31",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "Cybench",
      "score": 22.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Cybench leaderboard",
      "measured_at": "2025-01-31",
      "source": "Cybench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 7.36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-01-31",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "Aider polyglot",
      "score": 60.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-01-31",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "GSO-Bench",
      "score": 1.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2025-01-31",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "Fiction.LiveBench",
      "score": 50,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-01-31",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 19.58,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-01-31",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-01-31",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-01-31",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 34.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-01-31",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o3-mini",
      "model_slug": "o3-mini",
      "developer": "OpenAI",
      "benchmark": "Lech Mazur Writing",
      "score": 61.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-01-31",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 90.42,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-24",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "ProofBench",
      "score": 99,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-07-24",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 95.66,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 98.89,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 96.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-07-24",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 91.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-07-24",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 75.82,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-07-24",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 59.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-24",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 40.03,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-24",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 76.72,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-07-24",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "Balrog",
      "score": 63.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2026-07-24",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 90.12,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 73.17,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "EBR-bench",
      "score": 45.71,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-24",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "Mystery Game Puzzles",
      "score": 54.84,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "MirrorCode",
      "score": 87.78,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-24",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "PostTrainBench",
      "score": 35.66,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-24",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "FrontierCode",
      "score": 53.38,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-07-24",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "Furniture Assembly",
      "score": 44.05,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-24",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "FrontierSWE",
      "score": 52.01,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-07-24",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "WeirdML v3",
      "score": 36.15,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-07-24",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "OSWorld 2.0",
      "score": 31.43,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://osworld-v2.xlang.ai/",
      "measured_at": "2026-07-24",
      "source": "https://osworld-v2.xlang.ai/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 97.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-07-24",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 5",
      "model_slug": "claude-opus-5",
      "developer": "Anthropic",
      "benchmark": "APEX-Agents",
      "score": 65.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-24",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3-pro",
      "model_slug": "o3-pro",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 4.86,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-06-10",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3-pro",
      "model_slug": "o3-pro",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 78.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-06-10",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3-pro",
      "model_slug": "o3-pro",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 58.21,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-06-10",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3-pro",
      "model_slug": "o3-pro",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 45.27,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-06-10",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o3-pro",
      "model_slug": "o3-pro",
      "developer": "OpenAI",
      "benchmark": "Aider polyglot",
      "score": 84.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-06-10",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3-pro",
      "model_slug": "o3-pro",
      "developer": "OpenAI",
      "benchmark": "Fiction.LiveBench",
      "score": 97.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-06-10",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o3-pro",
      "model_slug": "o3-pro",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 59.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2025-06-10",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o3-pro",
      "model_slug": "o3-pro",
      "developer": "OpenAI",
      "benchmark": "Lech Mazur Writing",
      "score": 84.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-06-10",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 6.11,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 75.85,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 97.83,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 81.65,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 62.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-04-16",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "CadEval",
      "score": 62,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "CadEval Dashboard",
      "measured_at": "2025-04-16",
      "source": "CadEval Dashboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 52.56,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-04-16",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 31.18,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-04-16",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 19.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 17.93,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 26.44,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-04-16",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "METR Time Horizons",
      "score": 63.93,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-04-16",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "Aider polyglot",
      "score": 72,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-04-16",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "GeoBench",
      "score": 64,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2025-04-16",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "VPCT",
      "score": 36.25,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-04-16",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "GSO-Bench",
      "score": 3.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2025-04-16",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "Fiction.LiveBench",
      "score": 77.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-04-16",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 38.04,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 4.88,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-16",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "GDPval",
      "score": 25.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-16",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 58.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2025-04-16",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "o4-mini",
      "model_slug": "o4-mini",
      "developer": "OpenAI",
      "benchmark": "Lech Mazur Writing",
      "score": 75,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-04-16",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "open_llama_7b",
      "model_slug": "open-llama-7b",
      "developer": "OpenLM Research",
      "benchmark": "ARC AI2",
      "score": 18.27,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-06-07",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "open_llama_7b",
      "model_slug": "open-llama-7b",
      "developer": "OpenLM Research",
      "benchmark": "OpenBookQA",
      "score": 18.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-06-07",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "open_llama_7b",
      "model_slug": "open-llama-7b",
      "developer": "OpenLM Research",
      "benchmark": "Winogrande",
      "score": 34,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-06-07",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "open_llama_7b",
      "model_slug": "open-llama-7b",
      "developer": "OpenLM Research",
      "benchmark": "HellaSwag",
      "score": 62.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-06-07",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "open_llama_7b",
      "model_slug": "open-llama-7b",
      "developer": "OpenLM Research",
      "benchmark": "MMLU",
      "score": 6.53,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-06-07",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "open_llama_7b",
      "model_slug": "open-llama-7b",
      "developer": "OpenLM Research",
      "benchmark": "PIQA",
      "score": 52,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-06-07",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "phi-3-medium 14B",
      "model_slug": "phi-3-medium-14b",
      "developer": "Microsoft",
      "benchmark": "ARC AI2",
      "score": 88.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-medium 14B",
      "model_slug": "phi-3-medium-14b",
      "developer": "Microsoft",
      "benchmark": "OpenBookQA",
      "score": 83.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-medium 14B",
      "model_slug": "phi-3-medium-14b",
      "developer": "Microsoft",
      "benchmark": "Winogrande",
      "score": 63,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-medium 14B",
      "model_slug": "phi-3-medium-14b",
      "developer": "Microsoft",
      "benchmark": "GPQA diamond",
      "score": 3.59,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-04-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "phi-3-medium 14B",
      "model_slug": "phi-3-medium-14b",
      "developer": "Microsoft",
      "benchmark": "MATH level 5",
      "score": 17.56,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-04-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "phi-3-medium 14B",
      "model_slug": "phi-3-medium-14b",
      "developer": "Microsoft",
      "benchmark": "TriviaQA",
      "score": 73.9,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "phi-3-medium 14B",
      "model_slug": "phi-3-medium-14b",
      "developer": "Microsoft",
      "benchmark": "BBH",
      "score": 75.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-medium 14B",
      "model_slug": "phi-3-medium-14b",
      "developer": "Microsoft",
      "benchmark": "HellaSwag",
      "score": 76.53,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-medium 14B",
      "model_slug": "phi-3-medium-14b",
      "developer": "Microsoft",
      "benchmark": "ANLI",
      "score": 33.7,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "phi-3-medium 14B",
      "model_slug": "phi-3-medium-14b",
      "developer": "Microsoft",
      "benchmark": "MMLU",
      "score": 70.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-mini 3.8B",
      "model_slug": "phi-3-mini-3-8b",
      "developer": "Microsoft",
      "benchmark": "ARC AI2",
      "score": 79.87,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-mini 3.8B",
      "model_slug": "phi-3-mini-3-8b",
      "developer": "Microsoft",
      "benchmark": "OpenBookQA",
      "score": 84,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-mini 3.8B",
      "model_slug": "phi-3-mini-3-8b",
      "developer": "Microsoft",
      "benchmark": "Winogrande",
      "score": 41.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-mini 3.8B",
      "model_slug": "phi-3-mini-3-8b",
      "developer": "Microsoft",
      "benchmark": "TriviaQA",
      "score": 64,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "phi-3-mini 3.8B",
      "model_slug": "phi-3-mini-3-8b",
      "developer": "Microsoft",
      "benchmark": "BBH",
      "score": 62.27,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-mini 3.8B",
      "model_slug": "phi-3-mini-3-8b",
      "developer": "Microsoft",
      "benchmark": "HellaSwag",
      "score": 68.93,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-mini 3.8B",
      "model_slug": "phi-3-mini-3-8b",
      "developer": "Microsoft",
      "benchmark": "ANLI",
      "score": 29.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "phi-3-mini 3.8B",
      "model_slug": "phi-3-mini-3-8b",
      "developer": "Microsoft",
      "benchmark": "MMLU",
      "score": 58.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-small 7.4B",
      "model_slug": "phi-3-small-7-4b",
      "developer": "Microsoft",
      "benchmark": "ARC AI2",
      "score": 87.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-small 7.4B",
      "model_slug": "phi-3-small-7-4b",
      "developer": "Microsoft",
      "benchmark": "OpenBookQA",
      "score": 84,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-small 7.4B",
      "model_slug": "phi-3-small-7-4b",
      "developer": "Microsoft",
      "benchmark": "Winogrande",
      "score": 63,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-small 7.4B",
      "model_slug": "phi-3-small-7-4b",
      "developer": "Microsoft",
      "benchmark": "TriviaQA",
      "score": 58.1,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "phi-3-small 7.4B",
      "model_slug": "phi-3-small-7-4b",
      "developer": "Microsoft",
      "benchmark": "BBH",
      "score": 72.13,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your PhonePhi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your PhonePhi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-small 7.4B",
      "model_slug": "phi-3-small-7-4b",
      "developer": "Microsoft",
      "benchmark": "HellaSwag",
      "score": 69.33,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "phi-3-small 7.4B",
      "model_slug": "phi-3-small-7-4b",
      "developer": "Microsoft",
      "benchmark": "ANLI",
      "score": 37.15,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "phi-3-small 7.4B",
      "model_slug": "phi-3-small-7-4b",
      "developer": "Microsoft",
      "benchmark": "MMLU",
      "score": 67.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-04-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "vicuna-13b-v1.1",
      "model_slug": "vicuna-13b-v1-1",
      "developer": "LMSYS",
      "benchmark": "ARC AI2",
      "score": 24.27,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Textbooks Are All You Need II: phi-1.5 technical report",
      "measured_at": "2023-04-12",
      "source": "Textbooks Are All You Need II: phi-1.5 technical report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "vicuna-13b-v1.1",
      "model_slug": "vicuna-13b-v1-1",
      "developer": "LMSYS",
      "benchmark": "OpenBookQA",
      "score": 10.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Textbooks Are All You Need II: phi-1.5 technical report",
      "measured_at": "2023-04-12",
      "source": "Textbooks Are All You Need II: phi-1.5 technical report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "vicuna-13b-v1.1",
      "model_slug": "vicuna-13b-v1-1",
      "developer": "LMSYS",
      "benchmark": "Winogrande",
      "score": 41.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Textbooks Are All You Need II: phi-1.5 technical report",
      "measured_at": "2023-04-12",
      "source": "Textbooks Are All You Need II: phi-1.5 technical report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "vicuna-13b-v1.1",
      "model_slug": "vicuna-13b-v1-1",
      "developer": "LMSYS",
      "benchmark": "BBH",
      "score": 24.05,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Baichuan 2: Open Large-scale Language Models",
      "measured_at": "2023-04-12",
      "source": "Baichuan 2: Open Large-scale Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "vicuna-13b-v1.1",
      "model_slug": "vicuna-13b-v1-1",
      "developer": "LMSYS",
      "benchmark": "GSM8K",
      "score": 28.13,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Baichuan 2: Open Large-scale Language Models",
      "measured_at": "2023-04-12",
      "source": "Baichuan 2: Open Large-scale Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "vicuna-13b-v1.1",
      "model_slug": "vicuna-13b-v1-1",
      "developer": "LMSYS",
      "benchmark": "HellaSwag",
      "score": 43.73,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Textbooks Are All You Need II: phi-1.5 technical report",
      "measured_at": "2023-04-12",
      "source": "Textbooks Are All You Need II: phi-1.5 technical report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "vicuna-13b-v1.1",
      "model_slug": "vicuna-13b-v1-1",
      "developer": "LMSYS",
      "benchmark": "PIQA",
      "score": 54.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Textbooks Are All You Need II: phi-1.5 technical report",
      "measured_at": "2023-04-12",
      "source": "Textbooks Are All You Need II: phi-1.5 technical report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 93.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "ProofBench",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-09-22",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 91.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 98.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-09-22",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 80.27,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-09-22",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 72.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 65.28,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 86.08,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-09-22",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 96.03,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 95,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "EBR-bench",
      "score": 71.43,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "Mystery Game Puzzles",
      "score": 68.05,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "MirrorCode",
      "score": 95.56,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "PostTrainBench",
      "score": 43.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "Remote Labor Index",
      "score": 21.25,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "FrontierCode",
      "score": 54.64,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-09-22",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "Furniture Assembly",
      "score": 76.19,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "FrontierSWE",
      "score": 62.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-09-22",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "WeirdML v3",
      "score": 55.37,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-09-22",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "SpireClimb",
      "score": 5.83,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 98.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-09-22",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Opus 5.5",
      "model_slug": "claude-opus-5-5",
      "developer": "Anthropic",
      "benchmark": "APEX-Agents",
      "score": 73.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 5.93,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-05-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 75.26,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "MATH level 5",
      "score": 84.37,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 71.08,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 61.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-05-22",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 46.11,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-05-22",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 34.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-05-22",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "Cybench",
      "score": 35,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Cybench leaderboard",
      "measured_at": "2025-05-22",
      "source": "Cybench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 34.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-05-22",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "METR Time Horizons",
      "score": 61.97,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-05-22",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 7.26,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "Aider polyglot",
      "score": 61.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-05-22",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "GeoBench",
      "score": 37,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2025-05-22",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "VPCT",
      "score": 1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-05-22",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "GSO-Bench",
      "score": 4.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GSO Leaderboard",
      "measured_at": "2025-05-22",
      "source": "GSO Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "The Agent Company",
      "score": 33.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "TheAgentCompany leaderboard",
      "measured_at": "2025-05-22",
      "source": "TheAgentCompany leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "DeepResearch Bench",
      "score": 46.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2025-05-22",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "Fiction.LiveBench",
      "score": 46.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-05-22",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "OSWorld",
      "score": 43.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "OS World Website",
      "measured_at": "2025-05-22",
      "source": "OS World Website",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 40,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-05-22",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Sonnet 4",
      "model_slug": "claude-sonnet-4",
      "developer": "Anthropic",
      "benchmark": "Lech Mazur Writing",
      "score": 81.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-05-22",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 13.61,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "ProofBench",
      "score": 19,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2025-09-29",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 79.62,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-09-29",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "MATH level 5",
      "score": 97.73,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-09-29",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 77.76,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-09-29",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 72,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-09-29",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 47.71,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-09-29",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 45.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-09-29",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 30.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 5.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "Cybench",
      "score": 60,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Opus 4.5 System Card",
      "measured_at": "2025-09-29",
      "source": "Opus 4.5 System Card",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 45.16,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-09-29",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "METR Time Horizons",
      "score": 67.38,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-09-29",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "GeoBench",
      "score": 60,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2025-09-29",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "VPCT",
      "score": 9.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-09-29",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "GSO-Bench",
      "score": 14.71,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2025-09-29",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "DeepResearch Bench",
      "score": 52.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "DeepResearchBench Leaderboard",
      "measured_at": "2025-09-29",
      "source": "DeepResearchBench Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "OSWorld",
      "score": 62.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "OS World Website",
      "measured_at": "2025-09-29",
      "source": "OS World Website",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 25.12,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-09-29",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 2.44,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-09-29",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "EBR-bench",
      "score": 2.38,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-09-29",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "Mystery Game Puzzles",
      "score": 8.57,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-09-29",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "Remote Labor Index",
      "score": 2.08,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "Terminal Bench",
      "score": 46.52,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2025-09-29",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "GDPval",
      "score": 42.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 63.67,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Sonnet 4.5",
      "model_slug": "claude-sonnet-4-5",
      "developer": "Anthropic",
      "benchmark": "SWE-Bench verified",
      "score": 71.28,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-09-29",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 60.42,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "ProofBench",
      "score": 45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-02-17",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 86.63,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 85.78,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 83.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-02-17",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 66.07,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "WeirdML Leaderboard",
      "measured_at": "2026-02-17",
      "source": "WeirdML Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 54.72,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-02-17",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 35.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-17",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 18.98,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 56.84,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "GeoBench",
      "score": 73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2026-02-17",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 13.83,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "DeepResearch Bench",
      "score": 54.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2026-02-17",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "OSWorld",
      "score": 72.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "OS World Website",
      "measured_at": "2026-02-17",
      "source": "OS World Website",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "Mystery Game Puzzles",
      "score": 7.47,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "FrontierCode",
      "score": 24.31,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-02-17",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "Terminal Bench",
      "score": 53.43,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2026-02-17",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "ExploitBench",
      "score": 23.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-17",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "OSWorld 2.0",
      "score": 9.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://osworld-v2.xlang.ai/",
      "measured_at": "2026-02-17",
      "source": "https://osworld-v2.xlang.ai/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 86.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-17",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "APEX-Agents",
      "score": 43,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-02-17",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 4.6",
      "model_slug": "claude-sonnet-4-6",
      "developer": "Anthropic",
      "benchmark": "SWE-Bench verified",
      "score": 75.21,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Beam 1.0",
      "model_slug": "beam-1-0",
      "developer": null,
      "benchmark": "ARC-AGI-2",
      "score": 23.47,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Beam 1.0",
      "model_slug": "beam-1-0",
      "developer": null,
      "benchmark": "ProofBench",
      "score": 18,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-10-05",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Beam 1.0",
      "model_slug": "beam-1-0",
      "developer": null,
      "benchmark": "ARC-AGI",
      "score": 74,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-10-05",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Beam 1.0",
      "model_slug": "beam-1-0",
      "developer": null,
      "benchmark": "APEX-Agents",
      "score": 48.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "ProofBench",
      "score": 77,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-06-30",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 91.01,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-30",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 94.72,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-30",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 87.47,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-06-30",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 68.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-06-30",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 58.85,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-06-30",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 33.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-30",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 42.13,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-30",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 52.72,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2026-06-30",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "GeoBench",
      "score": 74,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2026-06-30",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "GSO-Bench",
      "score": 37.25,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2026-06-30",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 69.07,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-30",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 29.27,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-30",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "Mystery Game Puzzles",
      "score": 28.4,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-06-30",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "MirrorCode",
      "score": 63.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-30",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "Surface Evolver Bench",
      "score": 60,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-06-30",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "FrontierCode",
      "score": 42.73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-06-30",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5",
      "model_slug": "claude-sonnet-5",
      "developer": "Anthropic",
      "benchmark": "APEX-Agents",
      "score": 54.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-06-30",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI-2",
      "score": 83.89,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "ProofBench",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-09-28",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 98.03,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-28",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-28",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 95.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-09-28",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 80.31,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-09-28",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 46.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-28",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 56.86,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 71.08,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-09-28",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 93.44,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 80.49,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "Mystery Game Puzzles",
      "score": 61.45,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "MirrorCode",
      "score": 97.78,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "FrontierCode",
      "score": 52.09,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-09-28",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "Furniture Assembly",
      "score": 64.29,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "FrontierSWE",
      "score": 61.89,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-09-28",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "WeirdML v3",
      "score": 41.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-09-28",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "SpireClimb",
      "score": 1.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-28",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "ARC-AGI",
      "score": 96.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-09-28",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude Sonnet 5.5",
      "model_slug": "claude-sonnet-5-5",
      "developer": "Anthropic",
      "benchmark": "APEX-Agents",
      "score": 75.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-28",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "CodeQwen1.5-7B",
      "model_slug": "codeqwen1-5-7b",
      "developer": "Qwen",
      "benchmark": "ARC AI2",
      "score": 14.27,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-04-15",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "CodeQwen1.5-7B",
      "model_slug": "codeqwen1-5-7b",
      "developer": "Qwen",
      "benchmark": "Winogrande",
      "score": 19.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-04-15",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "CodeQwen1.5-7B",
      "model_slug": "codeqwen1-5-7b",
      "developer": "Qwen",
      "benchmark": "GSM8K",
      "score": 37.7,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-04-15",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "CodeQwen1.5-7B",
      "model_slug": "codeqwen1-5-7b",
      "developer": "Qwen",
      "benchmark": "MMLU",
      "score": 20.67,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-04-15",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek Coder 33B",
      "model_slug": "deepseek-coder-33b",
      "developer": "DeepSeek",
      "benchmark": "ARC AI2",
      "score": 22.93,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2023-11-02",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek Coder 33B",
      "model_slug": "deepseek-coder-33b",
      "developer": "DeepSeek",
      "benchmark": "Winogrande",
      "score": 24,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2023-11-02",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek Coder 33B",
      "model_slug": "deepseek-coder-33b",
      "developer": "DeepSeek",
      "benchmark": "GSM8K",
      "score": 35.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2023-11-02",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek Coder 33B",
      "model_slug": "deepseek-coder-33b",
      "developer": "DeepSeek",
      "benchmark": "MMLU",
      "score": 19.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2023-11-02",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek Coder 6.7B",
      "model_slug": "deepseek-coder-6-7b",
      "developer": "DeepSeek",
      "benchmark": "ARC AI2",
      "score": 15.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2023-11-02",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek Coder 6.7B",
      "model_slug": "deepseek-coder-6-7b",
      "developer": "DeepSeek",
      "benchmark": "Winogrande",
      "score": 15.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2023-11-02",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek Coder 6.7B",
      "model_slug": "deepseek-coder-6-7b",
      "developer": "DeepSeek",
      "benchmark": "GSM8K",
      "score": 21.3,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2023-11-02",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek Coder 6.7B",
      "model_slug": "deepseek-coder-6-7b",
      "developer": "DeepSeek",
      "benchmark": "MMLU",
      "score": 15.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2023-11-02",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek V4 Flash 0731",
      "model_slug": "deepseek-v4-flash-0731",
      "developer": "DeepSeek",
      "benchmark": "ARC-AGI-2",
      "score": 61.39,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-31",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Flash 0731",
      "model_slug": "deepseek-v4-flash-0731",
      "developer": "DeepSeek",
      "benchmark": "ProofBench",
      "score": 56,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-07-31",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4 Flash 0731",
      "model_slug": "deepseek-v4-flash-0731",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 91.72,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-31",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4 Flash 0731",
      "model_slug": "deepseek-v4-flash-0731",
      "developer": "DeepSeek",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 94.44,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-31",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4 Flash 0731",
      "model_slug": "deepseek-v4-flash-0731",
      "developer": "DeepSeek",
      "benchmark": "DTBench",
      "score": 84.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-07-31",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Flash 0731",
      "model_slug": "deepseek-v4-flash-0731",
      "developer": "DeepSeek",
      "benchmark": "WeirdML",
      "score": 62.96,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-07-31",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4 Flash 0731",
      "model_slug": "deepseek-v4-flash-0731",
      "developer": "DeepSeek",
      "benchmark": "LMCA",
      "score": 49.09,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-07-31",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Flash 0731",
      "model_slug": "deepseek-v4-flash-0731",
      "developer": "DeepSeek",
      "benchmark": "SimpleQA Verified",
      "score": 33.63,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-31",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4 Flash 0731",
      "model_slug": "deepseek-v4-flash-0731",
      "developer": "DeepSeek",
      "benchmark": "Chess Puzzles",
      "score": 28.45,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-31",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Flash 0731",
      "model_slug": "deepseek-v4-flash-0731",
      "developer": "DeepSeek",
      "benchmark": "SimpleBench",
      "score": 53.32,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-07-31",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Flash 0731",
      "model_slug": "deepseek-v4-flash-0731",
      "developer": "DeepSeek",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 60.57,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-31",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Flash 0731",
      "model_slug": "deepseek-v4-flash-0731",
      "developer": "DeepSeek",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 24.39,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-31",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Flash 0731",
      "model_slug": "deepseek-v4-flash-0731",
      "developer": "DeepSeek",
      "benchmark": "Mystery Game Puzzles",
      "score": 27.3,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-31",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Flash 0731",
      "model_slug": "deepseek-v4-flash-0731",
      "developer": "DeepSeek",
      "benchmark": "FrontierCode",
      "score": 18.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-07-31",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4 Flash 0731",
      "model_slug": "deepseek-v4-flash-0731",
      "developer": "DeepSeek",
      "benchmark": "ARC-AGI",
      "score": 89,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-07-31",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "ARC-AGI-2",
      "score": 61.25,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-13",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "ProofBench",
      "score": 50,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-08-13",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 92.59,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 98.61,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "DTBench",
      "score": 89.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-08-13",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "WeirdML",
      "score": 66.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-08-13",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "LMCA",
      "score": 53.47,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-08-13",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "SimpleQA Verified",
      "score": 52.91,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-13",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "Chess Puzzles",
      "score": 43.18,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-13",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 67.96,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 26.83,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "Mystery Game Puzzles",
      "score": 37.21,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-08-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "MirrorCode",
      "score": 33.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-13",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "FrontierCode",
      "score": 28.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-08-13",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "ARC-AGI",
      "score": 90.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-08-13",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek V4 Pro 0813",
      "model_slug": "deepseek-v4-pro-0813",
      "developer": "DeepSeek",
      "benchmark": "APEX-Agents",
      "score": 47.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-13",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "ARC-AGI-2",
      "score": 72.92,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "ProofBench",
      "score": 54,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-09-09",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 90.05,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 98.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "DTBench",
      "score": 83.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-09-09",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "LMCA",
      "score": 55.26,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-09-09",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "SimpleQA Verified",
      "score": 42.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "Chess Puzzles",
      "score": 45.29,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "SimpleBench",
      "score": 60.04,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-09-09",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 70.91,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 26.83,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "Mystery Game Puzzles",
      "score": 37.21,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "MirrorCode",
      "score": 64.44,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "Surface Evolver Bench",
      "score": 46.25,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-09-09",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "FrontierCode",
      "score": 43.87,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-09-09",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "Furniture Assembly",
      "score": 5.95,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "WeirdML v3",
      "score": 12.86,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-09-09",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "ARC-AGI",
      "score": 94.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-09-09",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek V4.1 Flash",
      "model_slug": "deepseek-v4-1-flash",
      "developer": "DeepSeek",
      "benchmark": "APEX-Agents",
      "score": 39.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4.1 Flash 1017",
      "model_slug": "deepseek-v4-1-flash-1017",
      "developer": "DeepSeek",
      "benchmark": "ARC-AGI-2",
      "score": 78.06,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4.1 Flash 1017",
      "model_slug": "deepseek-v4-1-flash-1017",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 94.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-17",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4.1 Flash 1017",
      "model_slug": "deepseek-v4-1-flash-1017",
      "developer": "DeepSeek",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 99.44,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-17",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4.1 Flash 1017",
      "model_slug": "deepseek-v4-1-flash-1017",
      "developer": "DeepSeek",
      "benchmark": "SimpleQA Verified",
      "score": 42.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-17",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4.1 Flash 1017",
      "model_slug": "deepseek-v4-1-flash-1017",
      "developer": "DeepSeek",
      "benchmark": "Chess Puzzles",
      "score": 43.18,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4.1 Flash 1017",
      "model_slug": "deepseek-v4-1-flash-1017",
      "developer": "DeepSeek",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 74.98,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4.1 Flash 1017",
      "model_slug": "deepseek-v4-1-flash-1017",
      "developer": "DeepSeek",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 46.34,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4.1 Flash 1017",
      "model_slug": "deepseek-v4-1-flash-1017",
      "developer": "DeepSeek",
      "benchmark": "Mystery Game Puzzles",
      "score": 43.82,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek V4.1 Flash 1017",
      "model_slug": "deepseek-v4-1-flash-1017",
      "developer": "DeepSeek",
      "benchmark": "WeirdML v3",
      "score": 22.14,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-10-17",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek V4.1 Flash 1017",
      "model_slug": "deepseek-v4-1-flash-1017",
      "developer": "DeepSeek",
      "benchmark": "ARC-AGI",
      "score": 94.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-10-17",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek V4.1 Flash 1017",
      "model_slug": "deepseek-v4-1-flash-1017",
      "developer": "DeepSeek",
      "benchmark": "APEX-Agents",
      "score": 69.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-17",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-Coder-V2-Lite-Base",
      "model_slug": "deepseek-coder-v2-lite-base",
      "developer": "DeepSeek",
      "benchmark": "ARC AI2",
      "score": 43.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-06-13",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-Coder-V2-Lite-Base",
      "model_slug": "deepseek-coder-v2-lite-base",
      "developer": "DeepSeek",
      "benchmark": "Winogrande",
      "score": 45.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-06-13",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-Coder-V2-Lite-Base",
      "model_slug": "deepseek-coder-v2-lite-base",
      "developer": "DeepSeek",
      "benchmark": "GSM8K",
      "score": 67.1,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-06-13",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-Coder-V2-Lite-Base",
      "model_slug": "deepseek-coder-v2-lite-base",
      "developer": "DeepSeek",
      "benchmark": "MMLU",
      "score": 47.33,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen2.5-Coder Technical Report",
      "measured_at": "2024-06-13",
      "source": "Qwen2.5-Coder Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Cerebras-GPT-13B",
      "model_slug": "cerebras-gpt-13b",
      "developer": "Cerebras",
      "benchmark": "ARC AI2",
      "score": 9.87,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-03-20",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Cerebras-GPT-13B",
      "model_slug": "cerebras-gpt-13b",
      "developer": "Cerebras",
      "benchmark": "OpenBookQA",
      "score": 14.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-03-20",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Cerebras-GPT-13B",
      "model_slug": "cerebras-gpt-13b",
      "developer": "Cerebras",
      "benchmark": "Winogrande",
      "score": 21.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-03-20",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Cerebras-GPT-13B",
      "model_slug": "cerebras-gpt-13b",
      "developer": "Cerebras",
      "benchmark": "HellaSwag",
      "score": 45.87,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-03-20",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Cerebras-GPT-13B",
      "model_slug": "cerebras-gpt-13b",
      "developer": "Cerebras",
      "benchmark": "MMLU",
      "score": 1.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-03-20",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Cerebras-GPT-13B",
      "model_slug": "cerebras-gpt-13b",
      "developer": "Cerebras",
      "benchmark": "PIQA",
      "score": 47,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-03-20",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "DeepSeek-R1 (May 2025)",
      "model_slug": "deepseek-r1-may-2025",
      "developer": "DeepSeek",
      "benchmark": "ARC-AGI-2",
      "score": 1.12,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-05-28",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-R1 (May 2025)",
      "model_slug": "deepseek-r1-may-2025",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 71.29,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-28",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1 (May 2025)",
      "model_slug": "deepseek-r1-may-2025",
      "developer": "DeepSeek",
      "benchmark": "MATH level 5",
      "score": 96.64,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-28",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1 (May 2025)",
      "model_slug": "deepseek-r1-may-2025",
      "developer": "DeepSeek",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 66.36,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-05-28",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1 (May 2025)",
      "model_slug": "deepseek-r1-may-2025",
      "developer": "DeepSeek",
      "benchmark": "WeirdML",
      "score": 41.63,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-05-28",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1 (May 2025)",
      "model_slug": "deepseek-r1-may-2025",
      "developer": "DeepSeek",
      "benchmark": "SimpleBench",
      "score": 28.96,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2025-05-28",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-R1 (May 2025)",
      "model_slug": "deepseek-r1-may-2025",
      "developer": "DeepSeek",
      "benchmark": "METR Time Horizons",
      "score": 53.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-05-28",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-R1 (May 2025)",
      "model_slug": "deepseek-r1-may-2025",
      "developer": "DeepSeek",
      "benchmark": "Aider polyglot",
      "score": 71.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-05-28",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1 (May 2025)",
      "model_slug": "deepseek-r1-may-2025",
      "developer": "DeepSeek",
      "benchmark": "DeepResearch Bench",
      "score": 35.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "DeepResearchBench Leaderboard",
      "measured_at": "2025-05-28",
      "source": "DeepResearchBench Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-R1 (May 2025)",
      "model_slug": "deepseek-r1-may-2025",
      "developer": "DeepSeek",
      "benchmark": "Fiction.LiveBench",
      "score": 75,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-05-28",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1 (May 2025)",
      "model_slug": "deepseek-r1-may-2025",
      "developer": "DeepSeek",
      "benchmark": "ARC-AGI",
      "score": 21.21,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2025-05-28",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-R1 (May 2025)",
      "model_slug": "deepseek-r1-may-2025",
      "developer": "DeepSeek",
      "benchmark": "Lech Mazur Writing",
      "score": 81.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-05-28",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V2 (MoE-236B, May 2024)",
      "model_slug": "deepseek-v2-moe-236b-may-2024",
      "developer": "DeepSeek",
      "benchmark": "ARC AI2",
      "score": 89.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-05-07",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V2 (MoE-236B, May 2024)",
      "model_slug": "deepseek-v2-moe-236b-may-2024",
      "developer": "DeepSeek",
      "benchmark": "Winogrande",
      "score": 72.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-05-07",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V2 (MoE-236B, May 2024)",
      "model_slug": "deepseek-v2-moe-236b-may-2024",
      "developer": "DeepSeek",
      "benchmark": "TriviaQA",
      "score": 80,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-05-07",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "DeepSeek-V2 (MoE-236B, May 2024)",
      "model_slug": "deepseek-v2-moe-236b-may-2024",
      "developer": "DeepSeek",
      "benchmark": "BBH",
      "score": 71.73,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-05-07",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V2 (MoE-236B, May 2024)",
      "model_slug": "deepseek-v2-moe-236b-may-2024",
      "developer": "DeepSeek",
      "benchmark": "HellaSwag",
      "score": 82.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-05-07",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V2 (MoE-236B, May 2024)",
      "model_slug": "deepseek-v2-moe-236b-may-2024",
      "developer": "DeepSeek",
      "benchmark": "MMLU",
      "score": 71.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-05-07",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V2 (MoE-236B, May 2024)",
      "model_slug": "deepseek-v2-moe-236b-may-2024",
      "developer": "DeepSeek",
      "benchmark": "PIQA",
      "score": 67.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek-V3 Technical Report",
      "measured_at": "2024-05-07",
      "source": "DeepSeek-V3 Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "DeepSeek-V3 (Mar 2025)",
      "model_slug": "deepseek-v3-mar-2025",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 59.19,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-03-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3 (Mar 2025)",
      "model_slug": "deepseek-v3-mar-2025",
      "developer": "DeepSeek",
      "benchmark": "MATH level 5",
      "score": 75.55,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-03-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3 (Mar 2025)",
      "model_slug": "deepseek-v3-mar-2025",
      "developer": "DeepSeek",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 37.72,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-03-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3 (Mar 2025)",
      "model_slug": "deepseek-v3-mar-2025",
      "developer": "DeepSeek",
      "benchmark": "DTBench",
      "score": 41.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-03-24",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V3 (Mar 2025)",
      "model_slug": "deepseek-v3-mar-2025",
      "developer": "DeepSeek",
      "benchmark": "WeirdML",
      "score": 36.08,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-03-24",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3 (Mar 2025)",
      "model_slug": "deepseek-v3-mar-2025",
      "developer": "DeepSeek",
      "benchmark": "LMCA",
      "score": 18.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-03-24",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V3 (Mar 2025)",
      "model_slug": "deepseek-v3-mar-2025",
      "developer": "DeepSeek",
      "benchmark": "SimpleBench",
      "score": 12.64,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-03-24",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V3 (Mar 2025)",
      "model_slug": "deepseek-v3-mar-2025",
      "developer": "DeepSeek",
      "benchmark": "METR Time Horizons",
      "score": 49.58,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-03-24",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V3 (Mar 2025)",
      "model_slug": "deepseek-v3-mar-2025",
      "developer": "DeepSeek",
      "benchmark": "Aider polyglot",
      "score": 55.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-03-24",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3 (Mar 2025)",
      "model_slug": "deepseek-v3-mar-2025",
      "developer": "DeepSeek",
      "benchmark": "Fiction.LiveBench",
      "score": 50,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-03-24",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3 (Mar 2025)",
      "model_slug": "deepseek-v3-mar-2025",
      "developer": "DeepSeek",
      "benchmark": "Lech Mazur Writing",
      "score": 77,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-03-24",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V3.1",
      "model_slug": "deepseek-v3-1",
      "developer": "DeepSeek",
      "benchmark": "DTBench",
      "score": 71.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-08-21",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V3.1",
      "model_slug": "deepseek-v3-1",
      "developer": "DeepSeek",
      "benchmark": "WeirdML",
      "score": 38.36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-08-21",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3.1",
      "model_slug": "deepseek-v3-1",
      "developer": "DeepSeek",
      "benchmark": "LMCA",
      "score": 28.58,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-08-21",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V3.1",
      "model_slug": "deepseek-v3-1",
      "developer": "DeepSeek",
      "benchmark": "SimpleBench",
      "score": 28,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2025-08-21",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V3.1",
      "model_slug": "deepseek-v3-1",
      "developer": "DeepSeek",
      "benchmark": "Fiction.LiveBench",
      "score": 52.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-08-21",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3.2-Exp",
      "model_slug": "deepseek-v3-2-exp",
      "developer": "DeepSeek",
      "benchmark": "DTBench",
      "score": 79.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-09-29",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V3.2-Exp",
      "model_slug": "deepseek-v3-2-exp",
      "developer": "DeepSeek",
      "benchmark": "WeirdML",
      "score": 39.48,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-09-29",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3.2-Exp",
      "model_slug": "deepseek-v3-2-exp",
      "developer": "DeepSeek",
      "benchmark": "LMCA",
      "score": 34.25,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-09-29",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V3.2-Exp",
      "model_slug": "deepseek-v3-2-exp",
      "developer": "DeepSeek",
      "benchmark": "Aider polyglot",
      "score": 74.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Aider LLM Leaderboards",
      "measured_at": "2025-09-29",
      "source": "Aider LLM Leaderboards",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3.2-Exp",
      "model_slug": "deepseek-v3-2-exp",
      "developer": "DeepSeek",
      "benchmark": "The Agent Company",
      "score": 42.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "TheAgentCompany leaderboard",
      "measured_at": "2025-09-29",
      "source": "TheAgentCompany leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3.2-Exp",
      "model_slug": "deepseek-v3-2-exp",
      "developer": "DeepSeek",
      "benchmark": "Fiction.LiveBench",
      "score": 83.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-09-29",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V3.2-Exp",
      "model_slug": "deepseek-v3-2-exp",
      "developer": "DeepSeek",
      "benchmark": "CL-bench",
      "score": 13.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-V3.2-Exp",
      "model_slug": "deepseek-v3-2-exp",
      "developer": "DeepSeek",
      "benchmark": "CL-bench Life",
      "score": 9.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V4-Flash",
      "model_slug": "deepseek-v4-flash",
      "developer": "DeepSeek",
      "benchmark": "DTBench",
      "score": 77.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-04-24",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V4-Flash",
      "model_slug": "deepseek-v4-flash",
      "developer": "DeepSeek",
      "benchmark": "WeirdML",
      "score": 45.63,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-04-24",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-V4-Flash",
      "model_slug": "deepseek-v4-flash",
      "developer": "DeepSeek",
      "benchmark": "LMCA",
      "score": 42.21,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-04-24",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "DeepSeek-V4-Flash",
      "model_slug": "deepseek-v4-flash",
      "developer": "DeepSeek",
      "benchmark": "SimpleBench",
      "score": 35.56,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-04-24",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 2",
      "model_slug": "claude-2",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 13.42,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2023-07-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 2",
      "model_slug": "claude-2",
      "developer": "Anthropic",
      "benchmark": "MATH level 5",
      "score": 11.73,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2023-07-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 2",
      "model_slug": "claude-2",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 2.4,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2023-07-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 2",
      "model_slug": "claude-2",
      "developer": "Anthropic",
      "benchmark": "TriviaQA",
      "score": 87.5,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Model Card and Evaluations for Claude Models",
      "measured_at": "2023-07-11",
      "source": "Model Card and Evaluations for Claude Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Claude 2",
      "model_slug": "claude-2",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 19.75,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2023-07-11",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 2",
      "model_slug": "claude-2",
      "developer": "Anthropic",
      "benchmark": "MMLU",
      "score": 71.33,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Model Card and Evaluations for Claude Models",
      "measured_at": "2023-07-11",
      "source": "Model Card and Evaluations for Claude Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Dolly 2.0-12b",
      "model_slug": "dolly-2-0-12b",
      "developer": "Databricks",
      "benchmark": "ARC AI2",
      "score": 19.47,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-04-11",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Dolly 2.0-12b",
      "model_slug": "dolly-2-0-12b",
      "developer": "Databricks",
      "benchmark": "OpenBookQA",
      "score": 18.93,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-04-11",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Dolly 2.0-12b",
      "model_slug": "dolly-2-0-12b",
      "developer": "Databricks",
      "benchmark": "Winogrande",
      "score": 23.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-04-11",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Dolly 2.0-12b",
      "model_slug": "dolly-2-0-12b",
      "developer": "Databricks",
      "benchmark": "HellaSwag",
      "score": 61.07,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-04-11",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Dolly 2.0-12b",
      "model_slug": "dolly-2-0-12b",
      "developer": "Databricks",
      "benchmark": "MMLU",
      "score": 1.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-04-11",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Dolly 2.0-12b",
      "model_slug": "dolly-2-0-12b",
      "developer": "Databricks",
      "benchmark": "PIQA",
      "score": 50.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "XGen-7B Technical Report",
      "measured_at": "2023-04-11",
      "source": "XGen-7B Technical Report",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Falcon 2 11B",
      "model_slug": "falcon-2-11b",
      "developer": "Technology Innovation Institute",
      "benchmark": "Winogrande",
      "score": 56.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Falcon2-11B Technical Report",
      "measured_at": "2024-05-09",
      "source": "Falcon2-11B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon 2 11B",
      "model_slug": "falcon-2-11b",
      "developer": "Technology Innovation Institute",
      "benchmark": "GSM8K",
      "score": 53.83,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Falcon2-11B Technical Report",
      "measured_at": "2024-05-09",
      "source": "Falcon2-11B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon 2 11B",
      "model_slug": "falcon-2-11b",
      "developer": "Technology Innovation Institute",
      "benchmark": "HellaSwag",
      "score": 77.21,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Falcon2-11B Technical Report",
      "measured_at": "2024-05-09",
      "source": "Falcon2-11B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon 2 11B",
      "model_slug": "falcon-2-11b",
      "developer": "Technology Innovation Institute",
      "benchmark": "MMLU",
      "score": 44.53,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Falcon2-11B Technical Report",
      "measured_at": "2024-05-09",
      "source": "Falcon2-11B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-180B",
      "model_slug": "falcon-180b",
      "developer": "Technology Innovation Institute",
      "benchmark": "ARC AI2",
      "score": 57.07,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-09-06",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-180B",
      "model_slug": "falcon-180b",
      "developer": "Technology Innovation Institute",
      "benchmark": "OpenBookQA",
      "score": 52.27,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-09-06",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-180B",
      "model_slug": "falcon-180b",
      "developer": "Technology Innovation Institute",
      "benchmark": "Winogrande",
      "score": 74.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-09-06",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-180B",
      "model_slug": "falcon-180b",
      "developer": "Technology Innovation Institute",
      "benchmark": "GSM8K",
      "score": 54.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Yi: Open Foundation Models by 01.AI",
      "measured_at": "2023-09-06",
      "source": "Yi: Open Foundation Models by 01.AI",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-180B",
      "model_slug": "falcon-180b",
      "developer": "Technology Innovation Institute",
      "benchmark": "HellaSwag",
      "score": 85.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-09-06",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-180B",
      "model_slug": "falcon-180b",
      "developer": "Technology Innovation Institute",
      "benchmark": "LAMBADA",
      "score": 79.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-09-06",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Falcon-180B",
      "model_slug": "falcon-180b",
      "developer": "Technology Innovation Institute",
      "benchmark": "MMLU",
      "score": 60.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-09-06",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-180B",
      "model_slug": "falcon-180b",
      "developer": "Technology Innovation Institute",
      "benchmark": "PIQA",
      "score": 69.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-09-06",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Falcon-40B",
      "model_slug": "falcon-40b",
      "developer": "Technology Innovation Institute",
      "benchmark": "ARC AI2",
      "score": 49.15,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Falcon2-11B Technical Report",
      "measured_at": "2023-03-15",
      "source": "Falcon2-11B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-40B",
      "model_slug": "falcon-40b",
      "developer": "Technology Innovation Institute",
      "benchmark": "OpenBookQA",
      "score": 42.13,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-03-15",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-40B",
      "model_slug": "falcon-40b",
      "developer": "Technology Innovation Institute",
      "benchmark": "Winogrande",
      "score": 53.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-03-15",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-40B",
      "model_slug": "falcon-40b",
      "developer": "Technology Innovation Institute",
      "benchmark": "TriviaQA",
      "score": 79.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "measured_at": "2023-03-15",
      "source": "Llama 2: Open Foundation and Fine-Tuned Chat Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Falcon-40B",
      "model_slug": "falcon-40b",
      "developer": "Technology Innovation Institute",
      "benchmark": "BBH",
      "score": 16.13,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Qwen Technical Report",
      "measured_at": "2023-03-15",
      "source": "Qwen Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-40B",
      "model_slug": "falcon-40b",
      "developer": "Technology Innovation Institute",
      "benchmark": "GSM8K",
      "score": 33.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford HELM",
      "measured_at": "2023-05-25",
      "source": "Stanford HELM",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-40B",
      "model_slug": "falcon-40b",
      "developer": "Technology Innovation Institute",
      "benchmark": "HellaSwag",
      "score": 80.37,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Falcon2-11B Technical Report",
      "measured_at": "2023-03-15",
      "source": "Falcon2-11B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-40B",
      "model_slug": "falcon-40b",
      "developer": "Technology Innovation Institute",
      "benchmark": "LAMBADA",
      "score": 77.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-03-15",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Falcon-40B",
      "model_slug": "falcon-40b",
      "developer": "Technology Innovation Institute",
      "benchmark": "MMLU",
      "score": 42.53,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Falcon2-11B Technical ReportFalcon2-11B Technical Report",
      "measured_at": "2023-03-15",
      "source": "Falcon2-11B Technical ReportFalcon2-11B Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Falcon-40B",
      "model_slug": "falcon-40b",
      "developer": "Technology Innovation Institute",
      "benchmark": "PIQA",
      "score": 66,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-03-15",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "GLM-4.7",
      "model_slug": "glm-4-7",
      "developer": "Zhipu AI",
      "benchmark": "ProofBench",
      "score": 6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2025-12-22",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-4.7",
      "model_slug": "glm-4-7",
      "developer": "Zhipu AI",
      "benchmark": "GPQA diamond",
      "score": 81.02,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-4.7",
      "model_slug": "glm-4-7",
      "developer": "Zhipu AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 83.32,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-4.7",
      "model_slug": "glm-4-7",
      "developer": "Zhipu AI",
      "benchmark": "SimpleQA Verified",
      "score": 32.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-4.7",
      "model_slug": "glm-4-7",
      "developer": "Zhipu AI",
      "benchmark": "Chess Puzzles",
      "score": 16.88,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-4.7",
      "model_slug": "glm-4-7",
      "developer": "Zhipu AI",
      "benchmark": "SimpleBench",
      "score": 37.24,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-12-22",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-4.7",
      "model_slug": "glm-4-7",
      "developer": "Zhipu AI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 4.28,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-4.7",
      "model_slug": "glm-4-7",
      "developer": "Zhipu AI",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-4.7",
      "model_slug": "glm-4-7",
      "developer": "Zhipu AI",
      "benchmark": "Terminal Bench",
      "score": 33.39,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2025-12-22",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-4.7",
      "model_slug": "glm-4-7",
      "developer": "Zhipu AI",
      "benchmark": "CL-bench",
      "score": 15.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-22",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GLM-4.7",
      "model_slug": "glm-4-7",
      "developer": "Zhipu AI",
      "benchmark": "CL-bench Life",
      "score": 10.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.1",
      "model_slug": "glm-5-1",
      "developer": "Zhipu AI",
      "benchmark": "ProofBench",
      "score": 22.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-04-07",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.1",
      "model_slug": "glm-5-1",
      "developer": "Zhipu AI",
      "benchmark": "GPQA diamond",
      "score": 90.14,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.1",
      "model_slug": "glm-5-1",
      "developer": "Zhipu AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 93.33,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.1",
      "model_slug": "glm-5-1",
      "developer": "Zhipu AI",
      "benchmark": "WeirdML",
      "score": 57.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-04-07",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.1",
      "model_slug": "glm-5-1",
      "developer": "Zhipu AI",
      "benchmark": "SimpleQA Verified",
      "score": 34,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-07",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.1",
      "model_slug": "glm-5-1",
      "developer": "Zhipu AI",
      "benchmark": "Chess Puzzles",
      "score": 15.82,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.1",
      "model_slug": "glm-5-1",
      "developer": "Zhipu AI",
      "benchmark": "SimpleBench",
      "score": 46.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2026-04-07",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.1",
      "model_slug": "glm-5-1",
      "developer": "Zhipu AI",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 20.83,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.1",
      "model_slug": "glm-5-1",
      "developer": "Zhipu AI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 38.78,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.1",
      "model_slug": "glm-5-1",
      "developer": "Zhipu AI",
      "benchmark": "ExploitBench",
      "score": 18.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-07",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.1",
      "model_slug": "glm-5-1",
      "developer": "Zhipu AI",
      "benchmark": "APEX-Agents",
      "score": 40.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-07",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.1",
      "model_slug": "glm-5-1",
      "developer": "Zhipu AI",
      "benchmark": "SWE-Bench verified",
      "score": 74.17,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "ProofBench",
      "score": 49,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-08-14",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "GPQA diamond",
      "score": 91.54,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-14",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 91.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-14",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "DTBench",
      "score": 79.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-08-14",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "WeirdML",
      "score": 75.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-08-14",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "LMCA",
      "score": 65.31,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-08-14",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "SimpleQA Verified",
      "score": 41,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-14",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "Chess Puzzles",
      "score": 33.71,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "SimpleBench",
      "score": 59.44,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-08-14",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 72.39,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 29.27,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "Mystery Game Puzzles",
      "score": 26.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "PostTrainBench",
      "score": 34.52,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-14",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "FrontierCode",
      "score": 40.14,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-08-14",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "FrontierSWE",
      "score": 30.18,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-08-14",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.3",
      "model_slug": "glm-5-3",
      "developer": "Zhipu AI",
      "benchmark": "APEX-Agents",
      "score": 56.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-14",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3 Haiku",
      "model_slug": "claude-3-haiku",
      "developer": "Anthropic",
      "benchmark": "Winogrande",
      "score": 48.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Claude 3 Model Family: Opus, Sonnet, Haiku",
      "measured_at": "2024-03-07",
      "source": "The Claude 3 Model Family: Opus, Sonnet, Haiku",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3 Haiku",
      "model_slug": "claude-3-haiku",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 15.7,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-03-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3 Haiku",
      "model_slug": "claude-3-haiku",
      "developer": "Anthropic",
      "benchmark": "MATH level 5",
      "score": 14.88,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-03-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3 Haiku",
      "model_slug": "claude-3-haiku",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 1.71,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-03-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3 Haiku",
      "model_slug": "claude-3-haiku",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 16.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-03-07",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3 Haiku",
      "model_slug": "claude-3-haiku",
      "developer": "Anthropic",
      "benchmark": "CadEval",
      "score": 12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "CadEval Dashboard",
      "measured_at": "2024-03-07",
      "source": "CadEval Dashboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3 Haiku",
      "model_slug": "claude-3-haiku",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 9.84,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-03-07",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3 Haiku",
      "model_slug": "claude-3-haiku",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 10.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2024-03-07",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3 Haiku",
      "model_slug": "claude-3-haiku",
      "developer": "Anthropic",
      "benchmark": "MMLU",
      "score": 65.07,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-03-07",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GLM-5.3-Flash",
      "model_slug": "glm-5-3-flash",
      "developer": "Zhipu AI",
      "benchmark": "ARC-AGI-2",
      "score": 65.83,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-20",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.3-Flash",
      "model_slug": "glm-5-3-flash",
      "developer": "Zhipu AI",
      "benchmark": "ProofBench",
      "score": 21,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-08-20",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.3-Flash",
      "model_slug": "glm-5-3-flash",
      "developer": "Zhipu AI",
      "benchmark": "GPQA diamond",
      "score": 90.49,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-20",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.3-Flash",
      "model_slug": "glm-5-3-flash",
      "developer": "Zhipu AI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 93.88,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-20",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.3-Flash",
      "model_slug": "glm-5-3-flash",
      "developer": "Zhipu AI",
      "benchmark": "Chess Puzzles",
      "score": 16.88,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-20",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.3-Flash",
      "model_slug": "glm-5-3-flash",
      "developer": "Zhipu AI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 58.73,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-20",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.3-Flash",
      "model_slug": "glm-5-3-flash",
      "developer": "Zhipu AI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 17.07,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-20",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.3-Flash",
      "model_slug": "glm-5-3-flash",
      "developer": "Zhipu AI",
      "benchmark": "Mystery Game Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-20",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GLM-5.3-Flash",
      "model_slug": "glm-5-3-flash",
      "developer": "Zhipu AI",
      "benchmark": "PostTrainBench",
      "score": 36.79,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-20",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GLM-5.3-Flash",
      "model_slug": "glm-5-3-flash",
      "developer": "Zhipu AI",
      "benchmark": "Surface Evolver Bench",
      "score": 52.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-08-20",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.3-Flash",
      "model_slug": "glm-5-3-flash",
      "developer": "Zhipu AI",
      "benchmark": "FrontierCode",
      "score": 31.83,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-08-20",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.3-Flash",
      "model_slug": "glm-5-3-flash",
      "developer": "Zhipu AI",
      "benchmark": "FrontierSWE",
      "score": 18.13,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-08-20",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GLM-5.3-Flash",
      "model_slug": "glm-5-3-flash",
      "developer": "Zhipu AI",
      "benchmark": "ARC-AGI",
      "score": 91,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-08-20",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GLM-5.3-Flash",
      "model_slug": "glm-5-3-flash",
      "developer": "Zhipu AI",
      "benchmark": "APEX-Agents",
      "score": 52.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-08-20",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-3.5 Turbo (Jan 2024)",
      "model_slug": "gpt-3-5-turbo-jan-2024",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 3.03,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-01-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-3.5 Turbo (Jan 2024)",
      "model_slug": "gpt-3-5-turbo-jan-2024",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 11.63,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-01-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-3.5 Turbo (Jan 2024)",
      "model_slug": "gpt-3-5-turbo-jan-2024",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 2.12,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-01-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-3.5 Turbo (Jan 2024)",
      "model_slug": "gpt-3-5-turbo-jan-2024",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 14.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-01-25",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-3.5 Turbo (Jan 2024)",
      "model_slug": "gpt-3-5-turbo-jan-2024",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 3.48,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-01-25",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-3.5 Turbo (Jan 2024)",
      "model_slug": "gpt-3-5-turbo-jan-2024",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 11.35,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2024-01-25",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-3.5 Turbo (Jan 2024)",
      "model_slug": "gpt-3-5-turbo-jan-2024",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-01-25",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-3.5 Turbo (Jan 2024)",
      "model_slug": "gpt-3-5-turbo-jan-2024",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-01-25",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-3.5 Turbo (Jan 2024)",
      "model_slug": "gpt-3-5-turbo-jan-2024",
      "developer": "OpenAI",
      "benchmark": "MMLU",
      "score": 56.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-01-25",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-3.5 Turbo (Jun 2023)",
      "model_slug": "gpt-3-5-turbo-jun-2023",
      "developer": "OpenAI",
      "benchmark": "Winogrande",
      "score": 63.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-06-13",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-3.5 Turbo (Jun 2023)",
      "model_slug": "gpt-3-5-turbo-jun-2023",
      "developer": "OpenAI",
      "benchmark": "BBH",
      "score": 48.79,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Baichuan 2: Open Large-scale Language Models",
      "measured_at": "2023-06-13",
      "source": "Baichuan 2: Open Large-scale Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-3.5 Turbo (Jun 2023)",
      "model_slug": "gpt-3-5-turbo-jun-2023",
      "developer": "OpenAI",
      "benchmark": "GSM8K",
      "score": 57.77,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Baichuan 2: Open Large-scale Language Models",
      "measured_at": "2023-06-13",
      "source": "Baichuan 2: Open Large-scale Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-3.5 Turbo (Jun 2023)",
      "model_slug": "gpt-3-5-turbo-jun-2023",
      "developer": "OpenAI",
      "benchmark": "MMLU",
      "score": 58.53,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2023-06-13",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-3.5 Turbo (Nov 2023)",
      "model_slug": "gpt-3-5-turbo-nov-2023",
      "developer": "OpenAI",
      "benchmark": "ARC AI2",
      "score": 83.2,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-11-06",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-3.5 Turbo (Nov 2023)",
      "model_slug": "gpt-3-5-turbo-nov-2023",
      "developer": "OpenAI",
      "benchmark": "OpenBookQA",
      "score": 81.33,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-11-06",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-3.5 Turbo (Nov 2023)",
      "model_slug": "gpt-3-5-turbo-nov-2023",
      "developer": "OpenAI",
      "benchmark": "Winogrande",
      "score": 37.6,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-11-06",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-3.5 Turbo (Nov 2023)",
      "model_slug": "gpt-3-5-turbo-nov-2023",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 4.21,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2023-11-06",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-3.5 Turbo (Nov 2023)",
      "model_slug": "gpt-3-5-turbo-nov-2023",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 15.89,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2023-11-06",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-3.5 Turbo (Nov 2023)",
      "model_slug": "gpt-3-5-turbo-nov-2023",
      "developer": "OpenAI",
      "benchmark": "TriviaQA",
      "score": 85.8,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-11-06",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "GPT-3.5 Turbo (Nov 2023)",
      "model_slug": "gpt-3-5-turbo-nov-2023",
      "developer": "OpenAI",
      "benchmark": "ANLI",
      "score": 37.15,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-11-06",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "GPT-3.5 Turbo (Nov 2023)",
      "model_slug": "gpt-3-5-turbo-nov-2023",
      "developer": "OpenAI",
      "benchmark": "MMLU",
      "score": 61.87,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2023-11-06",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4 (Jun 2023)",
      "model_slug": "gpt-4-jun-2023",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 7.85,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2023-06-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4 (Jun 2023)",
      "model_slug": "gpt-4-jun-2023",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 22.97,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2023-06-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4 (Jun 2023)",
      "model_slug": "gpt-4-jun-2023",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 1.01,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2023-06-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4 (Jun 2023)",
      "model_slug": "gpt-4-jun-2023",
      "developer": "OpenAI",
      "benchmark": "TriviaQA",
      "score": 84.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "PapersWithCode",
      "measured_at": "2023-06-13",
      "source": "PapersWithCode",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "GPT-4 (Jun 2023)",
      "model_slug": "gpt-4-jun-2023",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 37.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2023-06-13",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4 (Jun 2023)",
      "model_slug": "gpt-4-jun-2023",
      "developer": "OpenAI",
      "benchmark": "BBH",
      "score": 66.83,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Baichuan 2: Open Large-scale Language Models",
      "measured_at": "2023-06-13",
      "source": "Baichuan 2: Open Large-scale Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4 (Jun 2023)",
      "model_slug": "gpt-4-jun-2023",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 12.36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2023-06-13",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4 (Jun 2023)",
      "model_slug": "gpt-4-jun-2023",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 20.07,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2023-06-13",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4 (Jun 2023)",
      "model_slug": "gpt-4-jun-2023",
      "developer": "OpenAI",
      "benchmark": "GSM8K",
      "score": 89.99,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Baichuan 2: Open Large-scale Language Models",
      "measured_at": "2023-06-13",
      "source": "Baichuan 2: Open Large-scale Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4 (Jun 2023)",
      "model_slug": "gpt-4-jun-2023",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 3.06,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2023-06-13",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4 (Jun 2023)",
      "model_slug": "gpt-4-jun-2023",
      "developer": "OpenAI",
      "benchmark": "MMLU",
      "score": 76.53,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2023-06-13",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4 (Mar 2023)",
      "model_slug": "gpt-4-mar-2023",
      "developer": "OpenAI",
      "benchmark": "Winogrande",
      "score": 75,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "GPT-4 technical report",
      "measured_at": "2023-03-14",
      "source": "GPT-4 technical report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4 (Mar 2023)",
      "model_slug": "gpt-4-mar-2023",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 14.91,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2023-03-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4 (Mar 2023)",
      "model_slug": "gpt-4-mar-2023",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 0.46,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2023-03-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4 (Mar 2023)",
      "model_slug": "gpt-4-mar-2023",
      "developer": "OpenAI",
      "benchmark": "METR Time Horizons",
      "score": 36.11,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2023-03-14",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4 (Mar 2023)",
      "model_slug": "gpt-4-mar-2023",
      "developer": "OpenAI",
      "benchmark": "GSM8K",
      "score": 92,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "GPT-4 Technical Report",
      "measured_at": "2023-03-14",
      "source": "GPT-4 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4 (Mar 2023)",
      "model_slug": "gpt-4-mar-2023",
      "developer": "OpenAI",
      "benchmark": "HellaSwag",
      "score": 93.73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Falcon Series of Open Language Models",
      "measured_at": "2023-03-14",
      "source": "The Falcon Series of Open Language Models",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4 (Mar 2023)",
      "model_slug": "gpt-4-mar-2023",
      "developer": "OpenAI",
      "benchmark": "MMLU",
      "score": 81.87,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "GPT-4 Technical Report",
      "measured_at": "2023-03-14",
      "source": "GPT-4 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4 Turbo (Apr 2024)",
      "model_slug": "gpt-4-turbo-apr-2024",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 29.99,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-04-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4 Turbo (Apr 2024)",
      "model_slug": "gpt-4-turbo-apr-2024",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 46.73,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-04-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4 Turbo (Apr 2024)",
      "model_slug": "gpt-4-turbo-apr-2024",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 6.57,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-04-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4 Turbo (Apr 2024)",
      "model_slug": "gpt-4-turbo-apr-2024",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 18.01,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-04-09",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4 Turbo (Apr 2024)",
      "model_slug": "gpt-4-turbo-apr-2024",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 10.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2024-04-09",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4 Turbo (Apr 2024)",
      "model_slug": "gpt-4-turbo-apr-2024",
      "developer": "OpenAI",
      "benchmark": "METR Time Horizons",
      "score": 36.74,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2024-04-09",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4 Turbo (Apr 2024)",
      "model_slug": "gpt-4-turbo-apr-2024",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 0.74,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-04-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4 Turbo (Apr 2024)",
      "model_slug": "gpt-4-turbo-apr-2024",
      "developer": "OpenAI",
      "benchmark": "MMLU",
      "score": 75.07,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-04-09",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4 Turbo (Nov 2023)",
      "model_slug": "gpt-4-turbo-nov-2023",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 24.11,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-01-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4 Turbo (Nov 2023)",
      "model_slug": "gpt-4-turbo-nov-2023",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 40.02,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-01-25",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4 Turbo (Nov 2023)",
      "model_slug": "gpt-4-turbo-nov-2023",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2023-11-06",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4 Turbo (Nov 2023)",
      "model_slug": "gpt-4-turbo-nov-2023",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 11.51,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2023-11-06",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4 Turbo (Nov 2023)",
      "model_slug": "gpt-4-turbo-nov-2023",
      "developer": "OpenAI",
      "benchmark": "METR Time Horizons",
      "score": 35.17,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2024-01-25",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4 Turbo (Nov 2023)",
      "model_slug": "gpt-4-turbo-nov-2023",
      "developer": "OpenAI",
      "benchmark": "MMLU",
      "score": 72.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2023-11-06",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 0.42,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 58.22,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 83.01,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 38.27,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 47.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-04-14",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "CadEval",
      "score": 42,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "CadEval Dashboard",
      "measured_at": "2025-04-14",
      "source": "CadEval Dashboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 39.04,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-04-14",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 30.13,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-04-14",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 31.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-14",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 6.36,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 12.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-04-14",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "Aider polyglot",
      "score": 52.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-04-14",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "GeoBench",
      "score": 72,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2025-04-14",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "Fiction.LiveBench",
      "score": 63.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-04-14",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 6.28,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 5.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-04-14",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4.1",
      "model_slug": "gpt-4-1",
      "developer": "OpenAI",
      "benchmark": "SWE-Bench verified",
      "score": 48.54,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4.1 mini",
      "model_slug": "gpt-4-1-mini",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1 mini",
      "model_slug": "gpt-4-1-mini",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 56.73,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1 mini",
      "model_slug": "gpt-4-1-mini",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 87.29,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1 mini",
      "model_slug": "gpt-4-1-mini",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 44.67,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1 mini",
      "model_slug": "gpt-4-1-mini",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 48,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-04-14",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1 mini",
      "model_slug": "gpt-4-1-mini",
      "developer": "OpenAI",
      "benchmark": "CadEval",
      "score": 16,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "CadEval Dashboard",
      "measured_at": "2025-04-14",
      "source": "CadEval Dashboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1 mini",
      "model_slug": "gpt-4-1-mini",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 37.61,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-04-14",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1 mini",
      "model_slug": "gpt-4-1-mini",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 24.85,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-04-14",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1 mini",
      "model_slug": "gpt-4-1-mini",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 12.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-14",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1 mini",
      "model_slug": "gpt-4-1-mini",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1 mini",
      "model_slug": "gpt-4-1-mini",
      "developer": "OpenAI",
      "benchmark": "Aider polyglot",
      "score": 32.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-04-14",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1 mini",
      "model_slug": "gpt-4-1-mini",
      "developer": "OpenAI",
      "benchmark": "Fiction.LiveBench",
      "score": 44.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-04-14",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1 mini",
      "model_slug": "gpt-4-1-mini",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 7.02,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1 mini",
      "model_slug": "gpt-4-1-mini",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1 mini",
      "model_slug": "gpt-4-1-mini",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 3.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-04-14",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3 Opus",
      "model_slug": "claude-3-opus",
      "developer": "Anthropic",
      "benchmark": "Winogrande",
      "score": 77,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Claude 3 Model Family: Opus, Sonnet, Haiku",
      "measured_at": "2024-02-29",
      "source": "The Claude 3 Model Family: Opus, Sonnet, Haiku",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3 Opus",
      "model_slug": "claude-3-opus",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 30.78,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-02-29",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3 Opus",
      "model_slug": "claude-3-opus",
      "developer": "Anthropic",
      "benchmark": "MATH level 5",
      "score": 37.48,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-02-29",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3 Opus",
      "model_slug": "claude-3-opus",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 4.63,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-02-29",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3 Opus",
      "model_slug": "claude-3-opus",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-02-29",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3 Opus",
      "model_slug": "claude-3-opus",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 19.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-02-29",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3 Opus",
      "model_slug": "claude-3-opus",
      "developer": "Anthropic",
      "benchmark": "LMCA",
      "score": 20,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2024-02-29",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3 Opus",
      "model_slug": "claude-3-opus",
      "developer": "Anthropic",
      "benchmark": "SimpleQA Verified",
      "score": 12.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-02-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3 Opus",
      "model_slug": "claude-3-opus",
      "developer": "Anthropic",
      "benchmark": "Chess Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-02-29",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3 Opus",
      "model_slug": "claude-3-opus",
      "developer": "Anthropic",
      "benchmark": "Cybench",
      "score": 10,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Cybench leaderboard",
      "measured_at": "2024-02-29",
      "source": "Cybench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3 Opus",
      "model_slug": "claude-3-opus",
      "developer": "Anthropic",
      "benchmark": "SimpleBench",
      "score": 8.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2024-02-29",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3 Opus",
      "model_slug": "claude-3-opus",
      "developer": "Anthropic",
      "benchmark": "METR Time Horizons",
      "score": 29.48,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "measured_at": "2024-02-29",
      "source": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3 Opus",
      "model_slug": "claude-3-opus",
      "developer": "Anthropic",
      "benchmark": "MMLU",
      "score": 79.47,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-02-29",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4.1 nano",
      "model_slug": "gpt-4-1-nano",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1 nano",
      "model_slug": "gpt-4-1-nano",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 33.23,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1 nano",
      "model_slug": "gpt-4-1-nano",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 70,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1 nano",
      "model_slug": "gpt-4-1-nano",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 28.82,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1 nano",
      "model_slug": "gpt-4-1-nano",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 20.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-04-14",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1 nano",
      "model_slug": "gpt-4-1-nano",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 18.98,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-04-14",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1 nano",
      "model_slug": "gpt-4-1-nano",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 6.42,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-04-14",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1 nano",
      "model_slug": "gpt-4-1-nano",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-14",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1 nano",
      "model_slug": "gpt-4-1-nano",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-04-14",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1 nano",
      "model_slug": "gpt-4-1-nano",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 1.81,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-04-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.1 nano",
      "model_slug": "gpt-4-1-nano",
      "developer": "OpenAI",
      "benchmark": "Aider polyglot",
      "score": 8.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-04-14",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1 nano",
      "model_slug": "gpt-4-1-nano",
      "developer": "OpenAI",
      "benchmark": "Fiction.LiveBench",
      "score": 25,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-04-14",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.1 nano",
      "model_slug": "gpt-4-1-nano",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-04-14",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4.5",
      "model_slug": "gpt-4-5",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 0.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-02-27",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.5",
      "model_slug": "gpt-4-5",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 60.68,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-02-27",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.5",
      "model_slug": "gpt-4-5",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 78.63,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-02-27",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.5",
      "model_slug": "gpt-4-5",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 37.72,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-02-27",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.5",
      "model_slug": "gpt-4-5",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 39.37,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-02-27",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.5",
      "model_slug": "gpt-4-5",
      "developer": "OpenAI",
      "benchmark": "Cybench",
      "score": 17.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Cybench leaderboard",
      "measured_at": "2025-02-27",
      "source": "Cybench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.5",
      "model_slug": "gpt-4-5",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 21.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-02-27",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4.5",
      "model_slug": "gpt-4-5",
      "developer": "OpenAI",
      "benchmark": "Aider polyglot",
      "score": 44.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-02-27",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.5",
      "model_slug": "gpt-4-5",
      "developer": "OpenAI",
      "benchmark": "VPCT",
      "score": 17.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-02-27",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.5",
      "model_slug": "gpt-4-5",
      "developer": "OpenAI",
      "benchmark": "Fiction.LiveBench",
      "score": 63.9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-02-27",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4.5",
      "model_slug": "gpt-4-5",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 10.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-02-27",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4.5",
      "model_slug": "gpt-4-5",
      "developer": "OpenAI",
      "benchmark": "Lech Mazur Writing",
      "score": 75.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-02-27",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4o (Aug 2024)",
      "model_slug": "gpt-4o-aug-2024",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 33.63,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-08-06",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Aug 2024)",
      "model_slug": "gpt-4o-aug-2024",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 53.28,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-08-06",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Aug 2024)",
      "model_slug": "gpt-4o-aug-2024",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 6.3,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-08-06",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Aug 2024)",
      "model_slug": "gpt-4o-aug-2024",
      "developer": "OpenAI",
      "benchmark": "CadEval",
      "score": 26,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "CadEval Dashboard",
      "measured_at": "2024-08-06",
      "source": "CadEval Dashboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4o (Aug 2024)",
      "model_slug": "gpt-4o-aug-2024",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 26,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-08-06",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Aug 2024)",
      "model_slug": "gpt-4o-aug-2024",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 4.25,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-08-06",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4o (Aug 2024)",
      "model_slug": "gpt-4o-aug-2024",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 1.36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2024-08-06",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4o (Aug 2024)",
      "model_slug": "gpt-4o-aug-2024",
      "developer": "OpenAI",
      "benchmark": "Aider polyglot",
      "score": 23.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2024-08-06",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Aug 2024)",
      "model_slug": "gpt-4o-aug-2024",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 0.37,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-08-06",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4o (Aug 2024)",
      "model_slug": "gpt-4o-aug-2024",
      "developer": "OpenAI",
      "benchmark": "MMLU",
      "score": 79.07,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-08-06",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4o (May 2024)",
      "model_slug": "gpt-4o-may-2024",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 33.19,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-05-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (May 2024)",
      "model_slug": "gpt-4o-may-2024",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 51.05,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-05-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (May 2024)",
      "model_slug": "gpt-4o-may-2024",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 6.16,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-05-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (May 2024)",
      "model_slug": "gpt-4o-may-2024",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 40.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-05-13",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4o (May 2024)",
      "model_slug": "gpt-4o-may-2024",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 19.52,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2024-05-13",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4o (May 2024)",
      "model_slug": "gpt-4o-may-2024",
      "developer": "OpenAI",
      "benchmark": "METR Time Horizons",
      "score": 40.76,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2024-05-13",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4o (May 2024)",
      "model_slug": "gpt-4o-may-2024",
      "developer": "OpenAI",
      "benchmark": "Balrog",
      "score": 32.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2024-05-13",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (May 2024)",
      "model_slug": "gpt-4o-may-2024",
      "developer": "OpenAI",
      "benchmark": "MMLU",
      "score": 78.93,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-05-13",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-11-20",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 31.78,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-11-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 49.77,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-11-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 6.16,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-11-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 25.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-11-20",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "Cybench",
      "score": 12.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Cybench leaderboard",
      "measured_at": "2024-11-20",
      "source": "Cybench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 0.6,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-11-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "Aider polyglot",
      "score": 18.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2024-11-20",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "GeoBench",
      "score": 71,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2024-11-20",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "VPCT",
      "score": 10,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2024-11-20",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "GSO-Bench",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2024-11-20",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "The Agent Company",
      "score": 8.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "TheAgentCompany experiment results github",
      "measured_at": "2024-11-20",
      "source": "TheAgentCompany experiment results github",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "GDPval",
      "score": 9.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-11-20",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "MMLU",
      "score": 84.13,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-4 Technical Report",
      "measured_at": "2024-11-20",
      "source": "Phi-4 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 4.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2024-11-20",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "Lech Mazur Writing",
      "score": 81.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2024-11-20",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4o (Nov 2024)",
      "model_slug": "gpt-4o-nov-2024",
      "developer": "OpenAI",
      "benchmark": "SWE-Bench verified",
      "score": 30.99,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-11-20",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-07-18",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 17.67,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 52.63,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 6.85,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 24,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-07-18",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 11.76,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-07-18",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 12.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2024-07-18",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 8.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-07-18",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-07-18",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2024-07-18",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "Aider polyglot",
      "score": 3.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2024-07-18",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "Balrog",
      "score": 17.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2024-07-18",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "GSM8K",
      "score": 91.3,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-07-18",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "GeoBench",
      "score": 64,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2024-07-18",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "VPCT",
      "score": 1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2024-07-18",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 0.74,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-07-18",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 3.06,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2024-07-18",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "MMLU",
      "score": 75.73,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-4 Technical Report",
      "measured_at": "2024-07-18",
      "source": "Phi-4 Technical Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "PIQA",
      "score": 77.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-07-18",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "GPT-4o mini",
      "model_slug": "gpt-4o-mini",
      "developer": "OpenAI",
      "benchmark": "Lech Mazur Writing",
      "score": 67.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2024-07-18",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 9.86,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "ProofBench",
      "score": 18,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2025-08-07",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 84.96,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 98.13,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 91.38,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 84.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-08-07",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 60.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-08-07",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 47.05,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-08-07",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 50.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 32.66,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 48.04,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2025-08-07",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "METR Time Horizons",
      "score": 69.61,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "METR - Measuring AI Ability to Complete Long Tasks",
      "measured_at": "2025-08-07",
      "source": "METR - Measuring AI Ability to Complete Long Tasks",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "Aider polyglot",
      "score": 88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2025-08-07",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "Balrog",
      "score": 32.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2025-08-07",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "GeoBench",
      "score": 81,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2025-08-07",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "VPCT",
      "score": 49,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-08-07",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "GSO-Bench",
      "score": 6.86,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2025-08-07",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "DeepResearch Bench",
      "score": 49.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2025-08-07",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "Fiction.LiveBench",
      "score": 97.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-08-07",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 58.36,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 21.95,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "EBR-bench",
      "score": 12.7,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 15.18,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "MirrorCode",
      "score": 4.44,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "Remote Labor Index",
      "score": 1.67,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "Terminal Bench",
      "score": 49.61,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2025-08-07",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "GDPval",
      "score": 34.8,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 65.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2025-08-07",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "Lech Mazur Writing",
      "score": 86,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-08-07",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5",
      "model_slug": "gpt-5",
      "developer": "OpenAI",
      "benchmark": "SWE-Bench verified",
      "score": 73.55,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5 Pro",
      "model_slug": "gpt-5-pro",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 18.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-10-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 Pro",
      "model_slug": "gpt-5-pro",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 60.39,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-10-07",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 Pro",
      "model_slug": "gpt-5-pro",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 45.29,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-10-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 Pro",
      "model_slug": "gpt-5-pro",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 53.92,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-10-07",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 Pro",
      "model_slug": "gpt-5-pro",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 58.73,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-10-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 Pro",
      "model_slug": "gpt-5-pro",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 19.51,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-10-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 Pro",
      "model_slug": "gpt-5-pro",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 70.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-10-07",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 4.44,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "ProofBench",
      "score": 9,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2025-08-07",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 69.44,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 97.85,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 86.65,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 67.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-08-07",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 52.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-08-07",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 40.24,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-08-07",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 21.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 20.03,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "VPCT",
      "score": 10.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-08-07",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "Fiction.LiveBench",
      "score": 69.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-08-07",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 49.12,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 12.2,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 0.86,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "Terminal Bench",
      "score": 34.83,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2025-08-07",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 54.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "Lech Mazur Writing",
      "score": 83.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2025-08-07",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5 mini",
      "model_slug": "gpt-5-mini",
      "developer": "OpenAI",
      "benchmark": "SWE-Bench verified",
      "score": 64.67,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 2.61,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "ProofBench",
      "score": 12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2025-08-07",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 61.73,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "MATH level 5",
      "score": 95.24,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 81.09,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 37.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-08-07",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 38.06,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-08-07",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 9.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-08-07",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 11.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 17.93,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "VPCT",
      "score": 5.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-08-07",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "Fiction.LiveBench",
      "score": 44.4,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Fiction.live leaderboard",
      "measured_at": "2025-08-07",
      "source": "Fiction.live leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 21.05,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 2.44,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-08-07",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "Terminal Bench",
      "score": 21.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2025-08-07",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5 nano",
      "model_slug": "gpt-5-nano",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 20.71,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-08-07",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3 Sonnet",
      "model_slug": "claude-3-sonnet",
      "developer": "Anthropic",
      "benchmark": "Winogrande",
      "score": 50.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "The Claude 3 Model Family: Opus, Sonnet, Haiku",
      "measured_at": "2024-02-29",
      "source": "The Claude 3 Model Family: Opus, Sonnet, Haiku",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3 Sonnet",
      "model_slug": "claude-3-sonnet",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 21.66,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-02-29",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3 Sonnet",
      "model_slug": "claude-3-sonnet",
      "developer": "Anthropic",
      "benchmark": "MATH level 5",
      "score": 18.17,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-02-29",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3 Sonnet",
      "model_slug": "claude-3-sonnet",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 2.4,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-02-29",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3 Sonnet",
      "model_slug": "claude-3-sonnet",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 22.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-02-29",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3 Sonnet",
      "model_slug": "claude-3-sonnet",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 10.16,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-02-29",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3 Sonnet",
      "model_slug": "claude-3-sonnet",
      "developer": "Anthropic",
      "benchmark": "MMLU",
      "score": 67.87,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-02-29",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 17.64,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-13",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 86.98,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 88.6,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 83.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-11-13",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 60.77,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-11-13",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 51.64,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-11-13",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 48,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-13",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 38.97,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-13",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 43.84,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2025-11-13",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 54.45,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-2025-07-01-Private",
      "score": 20.83,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "VPCT",
      "score": 38.05,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-11-13",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "GSO-Bench",
      "score": 13.73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2025-11-13",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "DeepResearch Bench",
      "score": 42.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2025-11-13",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 10.77,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-13",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "Terminal Bench",
      "score": 47.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Terminal-Bench v2 Leaderboard",
      "measured_at": "2025-11-13",
      "source": "Terminal-Bench v2 Leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "CL-bench",
      "score": 23.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-13",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "CL-bench Life",
      "score": 17.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-11-13",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 72.83,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2025-11-13",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.1",
      "model_slug": "gpt-5-1",
      "developer": "OpenAI",
      "benchmark": "SWE-Bench verified",
      "score": 67.98,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-11-13",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 52.91,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-11",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "ProofBench",
      "score": 15,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2025-12-11",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 92.22,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 96.11,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 84.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2025-12-11",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 72.19,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-12-11",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 51.66,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2025-12-11",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 37.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-11",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 45.29,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-11",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 34.96,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2025-12-11",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "METR Time Horizons",
      "score": 75.29,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-11",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "VPCT",
      "score": 76,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "VPCT leaderboard",
      "measured_at": "2025-12-11",
      "source": "VPCT leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "GSO-Bench",
      "score": 27.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2025-12-11",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "DeepResearch Bench",
      "score": 41.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2025-12-11",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 70.95,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 31.7,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "EBR-bench",
      "score": 23.02,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 15.18,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-11",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "MirrorCode",
      "score": 32.22,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-11",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "Remote Labor Index",
      "score": 2.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-11",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "Furniture Assembly",
      "score": 11.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-11",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "WeirdML v3",
      "score": 10.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2025-12-11",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "Terminal Bench",
      "score": 64.94,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2025-12-11",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "GDPval",
      "score": 49.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-11",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "CL-bench",
      "score": 18.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-11",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 86.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "ARC Prize Leaderboard",
      "measured_at": "2025-12-11",
      "source": "ARC Prize Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.2",
      "model_slug": "gpt-5-2",
      "developer": "OpenAI",
      "benchmark": "SWE-Bench verified",
      "score": 73.76,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.2 Pro",
      "model_slug": "gpt-5-2-pro",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 54.16,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-11",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2 Pro",
      "model_slug": "gpt-5-2-pro",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 57.91,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2025-12-11",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2 Pro",
      "model_slug": "gpt-5-2-pro",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 48.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2025-12-11",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2 Pro",
      "model_slug": "gpt-5-2-pro",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 77.89,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2 Pro",
      "model_slug": "gpt-5-2-pro",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 46,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2025-12-11",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.2 Pro",
      "model_slug": "gpt-5-2-pro",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 90.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2025-12-11",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 73.95,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "ProofBench",
      "score": 56,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-03-05",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 94.86,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 97.78,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 90.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-03-05",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 77.7,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-03-05",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 61.14,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-03-05",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 45.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 45.29,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "METR Time Horizons",
      "score": 74.34,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "measured_at": "2026-03-05",
      "source": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "GSO-Bench",
      "score": 31.37,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2026-03-05",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "DeepResearch Bench",
      "score": 35.1,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2026-03-05",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 82.73,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 49,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "EBR-bench",
      "score": 25.4,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 30.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "MirrorCode",
      "score": 23.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "PostTrainBench",
      "score": 17.76,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-05",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "Furniture Assembly",
      "score": 10.71,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "Terminal Bench",
      "score": 81.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2026-03-05",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "CL-bench",
      "score": 27.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-05",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "CL-bench Life",
      "score": 21.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 93.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-03-05",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "APEX-Agents",
      "score": 52.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4",
      "model_slug": "gpt-5-4",
      "developer": "OpenAI",
      "benchmark": "SWE-Bench verified",
      "score": 76.86,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.4 Mini",
      "model_slug": "gpt-5-4-mini",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 18.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Mini",
      "model_slug": "gpt-5-4-mini",
      "developer": "OpenAI",
      "benchmark": "ProofBench",
      "score": 21,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-03-17",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4 Mini",
      "model_slug": "gpt-5-4-mini",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 85.93,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4 Mini",
      "model_slug": "gpt-5-4-mini",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 88.88,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4 Mini",
      "model_slug": "gpt-5-4-mini",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 66.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-03-17",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Mini",
      "model_slug": "gpt-5-4-mini",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 60.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-03-17",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4 Mini",
      "model_slug": "gpt-5-4-mini",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 47.99,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-03-17",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Mini",
      "model_slug": "gpt-5-4-mini",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 29.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-17",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4 Mini",
      "model_slug": "gpt-5-4-mini",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 18.98,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Mini",
      "model_slug": "gpt-5-4-mini",
      "developer": "OpenAI",
      "benchmark": "DeepResearch Bench",
      "score": 36.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2026-03-17",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.4 Mini",
      "model_slug": "gpt-5-4-mini",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 53.92,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Mini",
      "model_slug": "gpt-5-4-mini",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 9.76,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Mini",
      "model_slug": "gpt-5-4-mini",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 1.96,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Mini",
      "model_slug": "gpt-5-4-mini",
      "developer": "OpenAI",
      "benchmark": "FrontierCode",
      "score": 27.04,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-03-17",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4 Mini",
      "model_slug": "gpt-5-4-mini",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 63.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-03-17",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.4 Nano",
      "model_slug": "gpt-5-4-nano",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 5.69,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Nano",
      "model_slug": "gpt-5-4-nano",
      "developer": "OpenAI",
      "benchmark": "ProofBench",
      "score": 5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-03-17",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4 Nano",
      "model_slug": "gpt-5-4-nano",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 74.26,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4 Nano",
      "model_slug": "gpt-5-4-nano",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 87.77,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4 Nano",
      "model_slug": "gpt-5-4-nano",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 67.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-03-17",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Nano",
      "model_slug": "gpt-5-4-nano",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 49.23,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-03-17",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4 Nano",
      "model_slug": "gpt-5-4-nano",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 43.36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-03-17",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Nano",
      "model_slug": "gpt-5-4-nano",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 11.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-17",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4 Nano",
      "model_slug": "gpt-5-4-nano",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 42.13,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Nano",
      "model_slug": "gpt-5-4-nano",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 47.28,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Nano",
      "model_slug": "gpt-5-4-nano",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 12.2,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-17",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Nano",
      "model_slug": "gpt-5-4-nano",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-17",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Nano",
      "model_slug": "gpt-5-4-nano",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 51.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-03-17",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.4 Pro",
      "model_slug": "gpt-5-4-pro",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 83.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Pro",
      "model_slug": "gpt-5-4-pro",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 96.67,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4 Pro",
      "model_slug": "gpt-5-4-pro",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 57.44,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-03-05",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4 Pro",
      "model_slug": "gpt-5-4-pro",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 46.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-05",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.4 Pro",
      "model_slug": "gpt-5-4-pro",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 56.86,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-03-05",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Pro",
      "model_slug": "gpt-5-4-pro",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 68.92,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2026-03-05",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Pro",
      "model_slug": "gpt-5-4-pro",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 86.8,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Pro",
      "model_slug": "gpt-5-4-pro",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 58.54,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-03-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.4 Pro",
      "model_slug": "gpt-5-4-pro",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 94.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-03-05",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 85,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-23",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "ProofBench",
      "score": 50,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-04-23",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 95.84,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 93.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-04-23",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 84.91,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-04-23",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 63.85,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-04-23",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 63,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-23",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 61.07,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-23",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 62.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-04-23",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "GSO-Bench",
      "score": 40.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2026-04-23",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "DeepResearch Bench",
      "score": 54,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://evals.futuresearch.ai/#drb",
      "measured_at": "2026-04-23",
      "source": "https://evals.futuresearch.ai/#drb",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 89.75,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 72.5,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "EBR-bench",
      "score": 34.29,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 51.53,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-23",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "MirrorCode",
      "score": 23.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-23",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "PostTrainBench",
      "score": 23.87,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-23",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "Remote Labor Index",
      "score": 6.25,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-23",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "Surface Evolver Bench",
      "score": 88.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-04-23",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "FrontierCode",
      "score": 42.96,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-04-23",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "Furniture Assembly",
      "score": 20.24,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-23",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "Terminal Bench",
      "score": 84.72,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "measured_at": "2026-04-23",
      "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "ExploitBench",
      "score": 47.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-23",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "CL-bench Life",
      "score": 22.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-23",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "OSWorld 2.0",
      "score": 13,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://osworld-v2.xlang.ai/",
      "measured_at": "2026-04-23",
      "source": "https://osworld-v2.xlang.ai/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 95,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-04-23",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "APEX-Agents",
      "score": 55.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-23",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5",
      "model_slug": "gpt-5-5",
      "developer": "OpenAI",
      "benchmark": "SWE-Bench verified",
      "score": 80.58,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.5 Pro",
      "model_slug": "gpt-5-5-pro",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 84.58,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-23",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5 Pro",
      "model_slug": "gpt-5-5-pro",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 95.72,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5 Pro",
      "model_slug": "gpt-5-5-pro",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.5 Pro",
      "model_slug": "gpt-5-5-pro",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 93.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-04-23",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5 Pro",
      "model_slug": "gpt-5-5-pro",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 63.46,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-04-23",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5 Pro",
      "model_slug": "gpt-5-5-pro",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 60.02,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-04-23",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5 Pro",
      "model_slug": "gpt-5-5-pro",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 72.28,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-04-23",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5 Pro",
      "model_slug": "gpt-5-5-pro",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 92.34,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5 Pro",
      "model_slug": "gpt-5-5-pro",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 78.05,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-04-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.5 Pro",
      "model_slug": "gpt-5-5-pro",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 96.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-04-23",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 59.54,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "ProofBench",
      "score": 60,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-07-09",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 92.5,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 98.33,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 81.78,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-07-09",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 60.86,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-07-09",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 57.11,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-07-09",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 41,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 40.03,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 36.16,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2026-07-09",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "Balrog",
      "score": 45.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2026-07-09",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 86.43,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 60.98,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 12.98,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "MirrorCode",
      "score": 21.11,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "Surface Evolver Bench",
      "score": 61.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-07-09",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "FrontierCode",
      "score": 39.81,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-07-09",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "Furniture Assembly",
      "score": 17.86,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "WeirdML v3",
      "score": 10.38,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-07-09",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-07-09",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.6 Luna",
      "model_slug": "gpt-5-6-luna",
      "developer": "OpenAI",
      "benchmark": "APEX-Agents",
      "score": 43,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Haiku",
      "model_slug": "claude-3-5-haiku",
      "developer": "Anthropic",
      "benchmark": "GPQA diamond",
      "score": 18.24,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-10-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Haiku",
      "model_slug": "claude-3-5-haiku",
      "developer": "Anthropic",
      "benchmark": "MATH level 5",
      "score": 46.36,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-10-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Haiku",
      "model_slug": "claude-3-5-haiku",
      "developer": "Anthropic",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 4.21,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-10-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Haiku",
      "model_slug": "claude-3-5-haiku",
      "developer": "Anthropic",
      "benchmark": "DTBench",
      "score": 27.77,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-10-22",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3.5 Haiku",
      "model_slug": "claude-3-5-haiku",
      "developer": "Anthropic",
      "benchmark": "CadEval",
      "score": 32,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "CadEval Dashboard",
      "measured_at": "2024-10-22",
      "source": "CadEval Dashboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3.5 Haiku",
      "model_slug": "claude-3-5-haiku",
      "developer": "Anthropic",
      "benchmark": "WeirdML",
      "score": 30.73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-10-22",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Haiku",
      "model_slug": "claude-3-5-haiku",
      "developer": "Anthropic",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 0.6,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-10-22",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Claude 3.5 Haiku",
      "model_slug": "claude-3-5-haiku",
      "developer": "Anthropic",
      "benchmark": "Aider polyglot",
      "score": 28,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://aider.chat/docs/leaderboards/",
      "measured_at": "2024-10-22",
      "source": "https://aider.chat/docs/leaderboards/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Haiku",
      "model_slug": "claude-3-5-haiku",
      "developer": "Anthropic",
      "benchmark": "Balrog",
      "score": 19.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2024-10-22",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Haiku",
      "model_slug": "claude-3-5-haiku",
      "developer": "Anthropic",
      "benchmark": "GeoBench",
      "score": 34,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2024-10-22",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Claude 3.5 Haiku",
      "model_slug": "claude-3-5-haiku",
      "developer": "Anthropic",
      "benchmark": "MMLU",
      "score": 65.73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-10-22",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Claude 3.5 Haiku",
      "model_slug": "claude-3-5-haiku",
      "developer": "Anthropic",
      "benchmark": "Lech Mazur Writing",
      "score": 73.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2024-10-22",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 92.5,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "ProofBench",
      "score": 83,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-07-09",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 95.14,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 93.33,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-07-09",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 89.43,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-07-09",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 69.66,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-07-09",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 69.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 66.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 66.04,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://lmcouncil.ai/benchmarks",
      "measured_at": "2026-07-09",
      "source": "https://lmcouncil.ai/benchmarks",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "Balrog",
      "score": 60,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2026-07-09",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "GeoBench",
      "score": 73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://geobench.org/",
      "measured_at": "2026-07-09",
      "source": "https://geobench.org/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "GSO-Bench",
      "score": 76.47,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2026-07-09",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 93.81,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 82.93,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "EBR-bench",
      "score": 44.76,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 53.73,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "MirrorCode",
      "score": 20,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "PostTrainBench",
      "score": 31.01,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "Surface Evolver Bench",
      "score": 93.12,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-07-09",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "FrontierCode",
      "score": 47.49,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-07-09",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "Furniture Assembly",
      "score": 38.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "FrontierSWE",
      "score": 32.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-07-09",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "WeirdML v3",
      "score": 23.83,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-07-09",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "OSWorld 2.0",
      "score": 27.34,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://osworld-v2.xlang.ai/",
      "measured_at": "2026-07-09",
      "source": "https://osworld-v2.xlang.ai/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 97.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-07-09",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.6 Sol",
      "model_slug": "gpt-5-6-sol",
      "developer": "OpenAI",
      "benchmark": "APEX-Agents",
      "score": 51.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 83.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "ProofBench",
      "score": 74,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-07-09",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 94.87,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 99.72,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 88.88,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-07-09",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 78.27,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-07-09",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 64.65,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-07-09",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 43.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 44.23,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 38.68,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "SimpleBench Leaderboard",
      "measured_at": "2026-07-09",
      "source": "SimpleBench Leaderboard",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "Balrog",
      "score": 53.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2026-07-09",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 90.49,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 70.73,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-07-09",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 28.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "MirrorCode",
      "score": 41.11,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "Surface Evolver Bench",
      "score": 83.75,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "measured_at": "2026-07-09",
      "source": "https://yhenon.github.io/surface-evolver-llm-eval/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "FrontierCode",
      "score": 41.31,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-07-09",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "Furniture Assembly",
      "score": 34.52,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 96.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-07-09",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.6 Terra",
      "model_slug": "gpt-5-6-terra",
      "developer": "OpenAI",
      "benchmark": "APEX-Agents",
      "score": 58.2,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-07-09",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 95,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-03",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "ProofBench",
      "score": 99,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-09-03",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 98.29,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-03",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-03",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 95.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-09-03",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "WeirdML",
      "score": 93.57,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-09-03",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 75.73,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-09-03",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 75.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-03",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 74.75,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-03",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 83.8,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-09-03",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "Balrog",
      "score": 68.3,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2026-09-03",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "GSO-Bench",
      "score": 79.41,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://gso-bench.github.io/index.html",
      "measured_at": "2026-09-03",
      "source": "https://gso-bench.github.io/index.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 98.61,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-03",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 97.6,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-03",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "EBR-bench",
      "score": 76.19,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-03",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 82.37,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-03",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "MirrorCode",
      "score": 60,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-03",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "PostTrainBench",
      "score": 41.88,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-03",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "Remote Labor Index",
      "score": 20.83,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-03",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "FrontierCode",
      "score": 53.26,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-09-03",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "Furniture Assembly",
      "score": 71.43,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-03",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "FrontierSWE",
      "score": 65.51,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.frontierswe.com/",
      "measured_at": "2026-09-03",
      "source": "https://www.frontierswe.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "WeirdML v3",
      "score": 60.05,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-09-03",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "SpireClimb",
      "score": 34,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-03",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 98.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-09-03",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-6 Astra",
      "model_slug": "gpt-6-astra",
      "developer": "OpenAI",
      "benchmark": "APEX-Agents",
      "score": 64.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-03",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 59.31,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "ProofBench",
      "score": 64,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-09-22",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 90.93,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 98.89,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 83.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-09-22",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 52.34,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-09-22",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 41.4,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 32.66,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 83.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 56.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "MirrorCode",
      "score": 26.67,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "FrontierCode",
      "score": 42.42,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-09-22",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "Furniture Assembly",
      "score": 20.24,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "WeirdML v3",
      "score": 13.36,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-09-22",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "SpireClimb",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 86.67,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-09-22",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-6 Luna",
      "model_slug": "gpt-6-luna",
      "developer": "OpenAI",
      "benchmark": "APEX-Agents",
      "score": 44.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 89.58,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "ProofBench",
      "score": 83,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-09-22",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 96.28,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 95.55,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-09-22",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 69.51,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-09-22",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 60.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 53.7,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 67.72,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-09-22",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 94.55,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 90,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "EBR-bench",
      "score": 53.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 51.53,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "MirrorCode",
      "score": 48.89,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "FrontierCode",
      "score": 49.27,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-09-22",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "Furniture Assembly",
      "score": 40.48,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "WeirdML v3",
      "score": 30.63,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-09-22",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "SpireClimb",
      "score": 9.17,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 95.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-09-22",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-6 Sol",
      "model_slug": "gpt-6-sol",
      "developer": "OpenAI",
      "benchmark": "APEX-Agents",
      "score": 54.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-22",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI-2",
      "score": 94.17,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "ProofBench",
      "score": 99,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://www.vals.ai/benchmarks/proof_bench",
      "measured_at": "2026-09-29",
      "source": "https://www.vals.ai/benchmarks/proof_bench",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 97.77,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "DTBench",
      "score": 94.22,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2026-09-29",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "LMCA",
      "score": 73.86,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/lmca",
      "measured_at": "2026-09-29",
      "source": "https://conceptualreasoning.ai/lmca",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 73.9,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "Chess Puzzles",
      "score": 66.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 79.48,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-09-29",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 98.61,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "EBR-bench",
      "score": 54.29,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 77.97,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "MirrorCode",
      "score": 61.11,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "PostTrainBench",
      "score": 38.79,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "FrontierCode",
      "score": 50.23,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://cognition.com/frontiercode",
      "measured_at": "2026-09-29",
      "source": "https://cognition.com/frontiercode",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "Furniture Assembly",
      "score": 71.43,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "WeirdML v3",
      "score": 56.23,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-09-29",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "SpireClimb",
      "score": 23.83,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "ARC-AGI",
      "score": 98.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://arcprize.org/leaderboard",
      "measured_at": "2026-09-29",
      "source": "https://arcprize.org/leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-6.1 Sol",
      "model_slug": "gpt-6-1-sol",
      "developer": "OpenAI",
      "benchmark": "APEX-Agents",
      "score": 60,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-09-29",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6.1 Terra",
      "model_slug": "gpt-6-1-terra",
      "developer": "OpenAI",
      "benchmark": "GPQA diamond",
      "score": 96.28,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-26",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6.1 Terra",
      "model_slug": "gpt-6-1-terra",
      "developer": "OpenAI",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 100,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-26",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6.1 Terra",
      "model_slug": "gpt-6-1-terra",
      "developer": "OpenAI",
      "benchmark": "SimpleQA Verified",
      "score": 57.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-26",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6.1 Terra",
      "model_slug": "gpt-6-1-terra",
      "developer": "OpenAI",
      "benchmark": "SimpleBench",
      "score": 66.64,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://simple-bench.com/",
      "measured_at": "2026-10-26",
      "source": "https://simple-bench.com/",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Terra",
      "model_slug": "gpt-6-1-terra",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tiers-1-3-v2-Private",
      "score": 95.66,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-26",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Terra",
      "model_slug": "gpt-6-1-terra",
      "developer": "OpenAI",
      "benchmark": "FrontierMath-Tier-4-v2-Private",
      "score": 92.68,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-26",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Terra",
      "model_slug": "gpt-6-1-terra",
      "developer": "OpenAI",
      "benchmark": "EBR-bench",
      "score": 78.1,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-26",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Terra",
      "model_slug": "gpt-6-1-terra",
      "developer": "OpenAI",
      "benchmark": "Mystery Game Puzzles",
      "score": 71.36,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-26",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Terra",
      "model_slug": "gpt-6-1-terra",
      "developer": "OpenAI",
      "benchmark": "MirrorCode",
      "score": 76.67,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-26",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Terra",
      "model_slug": "gpt-6-1-terra",
      "developer": "OpenAI",
      "benchmark": "Furniture Assembly",
      "score": 70.24,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-26",
      "source": "unknown",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "GPT-6.1 Terra",
      "model_slug": "gpt-6-1-terra",
      "developer": "OpenAI",
      "benchmark": "WeirdML v3",
      "score": 44.94,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2026-10-26",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6.1 Terra",
      "model_slug": "gpt-6-1-terra",
      "developer": "OpenAI",
      "benchmark": "SpireClimb",
      "score": 22.33,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-26",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "GPT-6.1 Terra",
      "model_slug": "gpt-6-1-terra",
      "developer": "OpenAI",
      "benchmark": "APEX-Agents",
      "score": 68.3,
      "metric": "accuracy (%)",
      "evidence": "unverified",
      "evaluator": null,
      "measured_at": "2026-10-26",
      "source": "unknown",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Flash (May 2024)",
      "model_slug": "gemini-1-5-flash-may-2024",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 21.35,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-05-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Flash (May 2024)",
      "model_slug": "gemini-1-5-flash-may-2024",
      "developer": "Google DeepMind",
      "benchmark": "MATH level 5",
      "score": 25.12,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-05-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Flash (May 2024)",
      "model_slug": "gemini-1-5-flash-may-2024",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 3.79,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-05-23",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Flash (May 2024)",
      "model_slug": "gemini-1-5-flash-may-2024",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 18.97,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-05-23",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 1.5 Flash (May 2024)",
      "model_slug": "gemini-1-5-flash-may-2024",
      "developer": "Google DeepMind",
      "benchmark": "GSM8K",
      "score": 82.4,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-05-23",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 1.5 Flash (May 2024)",
      "model_slug": "gemini-1-5-flash-may-2024",
      "developer": "Google DeepMind",
      "benchmark": "MMLU",
      "score": 70.53,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-05-23",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 1.5 Flash (Sep 2024)",
      "model_slug": "gemini-1-5-flash-sep-2024",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 31,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Flash (Sep 2024)",
      "model_slug": "gemini-1-5-flash-sep-2024",
      "developer": "Google DeepMind",
      "benchmark": "MATH level 5",
      "score": 61.87,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Flash (Sep 2024)",
      "model_slug": "gemini-1-5-flash-sep-2024",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 16.17,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Flash (Sep 2024)",
      "model_slug": "gemini-1-5-flash-sep-2024",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 22.98,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-09-24",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 1.5 Flash (Sep 2024)",
      "model_slug": "gemini-1-5-flash-sep-2024",
      "developer": "Google DeepMind",
      "benchmark": "WeirdML",
      "score": 24.87,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://htihle.github.io/weirdml.html",
      "measured_at": "2024-09-24",
      "source": "https://htihle.github.io/weirdml.html",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Flash (Sep 2024)",
      "model_slug": "gemini-1-5-flash-sep-2024",
      "developer": "Google DeepMind",
      "benchmark": "FrontierMath-2025-02-28-Private",
      "score": 0,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-09-24",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 1.5 Flash (Sep 2024)",
      "model_slug": "gemini-1-5-flash-sep-2024",
      "developer": "Google DeepMind",
      "benchmark": "Balrog",
      "score": 14.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://balrogai.com/",
      "measured_at": "2024-09-24",
      "source": "https://balrogai.com/",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Flash (Sep 2024)",
      "model_slug": "gemini-1-5-flash-sep-2024",
      "developer": "Google DeepMind",
      "benchmark": "GeoBench",
      "score": 76,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "GeoBench leaderboard",
      "measured_at": "2024-09-24",
      "source": "GeoBench leaderboard",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Flash (Sep 2024)",
      "model_slug": "gemini-1-5-flash-sep-2024",
      "developer": "Google DeepMind",
      "benchmark": "MMLU",
      "score": 65.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Stanford CRFM Leaderboard",
      "measured_at": "2024-09-24",
      "source": "Stanford CRFM Leaderboard",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 1.5 Flash (Sep 2024)",
      "model_slug": "gemini-1-5-flash-sep-2024",
      "developer": "Google DeepMind",
      "benchmark": "PIQA",
      "score": 75,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "measured_at": "2024-09-24",
      "source": "Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone",
      "benchmark_integrity_grade": "D"
    },
    {
      "model": "Gemini 1.5 Pro (May 2024)",
      "model_slug": "gemini-1-5-pro-may-2024",
      "developer": "Google DeepMind",
      "benchmark": "GPQA diamond",
      "score": 28.98,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-05-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Pro (May 2024)",
      "model_slug": "gemini-1-5-pro-may-2024",
      "developer": "Google DeepMind",
      "benchmark": "MATH level 5",
      "score": 40.75,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-05-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Pro (May 2024)",
      "model_slug": "gemini-1-5-pro-may-2024",
      "developer": "Google DeepMind",
      "benchmark": "OTIS Mock AIME 2024-2025",
      "score": 6.71,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2024-05-14",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "Gemini 1.5 Pro (May 2024)",
      "model_slug": "gemini-1-5-pro-may-2024",
      "developer": "Google DeepMind",
      "benchmark": "DTBench",
      "score": 27.45,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "https://conceptualreasoning.ai/dtbench",
      "measured_at": "2024-05-14",
      "source": "https://conceptualreasoning.ai/dtbench",
      "benchmark_integrity_grade": "A"
    },
    {
      "model": "Gemini 1.5 Pro (May 2024)",
      "model_slug": "gemini-1-5-pro-may-2024",
      "developer": "Google DeepMind",
      "benchmark": "BBH",
      "score": 85.6,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemini 1.5 report",
      "measured_at": "2024-05-14",
      "source": "Gemini 1.5 report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 1.5 Pro (May 2024)",
      "model_slug": "gemini-1-5-pro-may-2024",
      "developer": "Google DeepMind",
      "benchmark": "MMLU",
      "score": 81.2,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "Gemini 1.5 Report",
      "measured_at": "2024-05-14",
      "source": "Gemini 1.5 Report",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "Gemini 2.0 Flash (Dec 2024)",
      "model_slug": "gemini-2-0-flash-dec-2024",
      "developer": "Google DeepMind",
      "benchmark": "Lech Mazur Writing",
      "score": 71.5,
      "metric": "accuracy (%)",
      "evidence": "third_party",
      "evaluator": "lechmazur/writing Github repository",
      "measured_at": "2024-12-11",
      "source": "lechmazur/writing Github repository",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "GPT-5.3 Codex",
      "model_slug": "gpt-5-3-codex",
      "developer": "OpenAI",
      "benchmark": "SWE-Bench verified",
      "score": 74.79,
      "metric": "accuracy (%)",
      "evidence": "independently_reproduced",
      "evaluator": "Epoch AI",
      "measured_at": "2026-02-05",
      "source": "Epoch evaluations",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-7B",
      "model_slug": "deepseek-r1-distill-qwen-7b",
      "developer": "DeepSeek",
      "benchmark": "AIME",
      "score": 55.5,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-7B",
      "model_slug": "deepseek-r1-distill-qwen-7b",
      "developer": "DeepSeek",
      "benchmark": "Codeforces rating",
      "score": 1189,
      "metric": "rating",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": null
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-7B",
      "model_slug": "deepseek-r1-distill-qwen-7b",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 49.1,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-7B",
      "model_slug": "deepseek-r1-distill-qwen-7b",
      "developer": "DeepSeek",
      "benchmark": "LiveCodeBench",
      "score": 37.6,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-7B",
      "model_slug": "deepseek-r1-distill-qwen-7b",
      "developer": "DeepSeek",
      "benchmark": "MATH-500",
      "score": 92.8,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-14B",
      "model_slug": "deepseek-r1-distill-qwen-14b",
      "developer": "DeepSeek",
      "benchmark": "AIME",
      "score": 69.7,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-14B",
      "model_slug": "deepseek-r1-distill-qwen-14b",
      "developer": "DeepSeek",
      "benchmark": "Codeforces rating",
      "score": 1481,
      "metric": "rating",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": null
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-14B",
      "model_slug": "deepseek-r1-distill-qwen-14b",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 59.1,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-14B",
      "model_slug": "deepseek-r1-distill-qwen-14b",
      "developer": "DeepSeek",
      "benchmark": "LiveCodeBench",
      "score": 53.1,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Qwen-14B",
      "model_slug": "deepseek-r1-distill-qwen-14b",
      "developer": "DeepSeek",
      "benchmark": "MATH-500",
      "score": 93.9,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "C"
    },
    {
      "model": "DeepSeek-R1-Distill-Llama-8B",
      "model_slug": "deepseek-r1-distill-llama-8b",
      "developer": "DeepSeek",
      "benchmark": "AIME",
      "score": 50.4,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Llama-8B",
      "model_slug": "deepseek-r1-distill-llama-8b",
      "developer": "DeepSeek",
      "benchmark": "Codeforces rating",
      "score": 1205,
      "metric": "rating",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": null
    },
    {
      "model": "DeepSeek-R1-Distill-Llama-8B",
      "model_slug": "deepseek-r1-distill-llama-8b",
      "developer": "DeepSeek",
      "benchmark": "GPQA diamond",
      "score": 49,
      "metric": "accuracy (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Llama-8B",
      "model_slug": "deepseek-r1-distill-llama-8b",
      "developer": "DeepSeek",
      "benchmark": "LiveCodeBench",
      "score": 39.6,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "B"
    },
    {
      "model": "DeepSeek-R1-Distill-Llama-8B",
      "model_slug": "deepseek-r1-distill-llama-8b",
      "developer": "DeepSeek",
      "benchmark": "MATH-500",
      "score": 89.1,
      "metric": "pass@1 (%)",
      "evidence": "self_reported",
      "evaluator": "DeepSeek AI",
      "measured_at": "2025-01-22",
      "source": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (arXiv:2501.12948), Table 5",
      "benchmark_integrity_grade": "C"
    }
  ]
}
