Anthropic·released 2025-08-051 source
Claude Opus 4.1 benchmark scores: 20 benchmarks tracked. 40% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 0 of 20 benchmarks·best result 84.7 on Lech Mazur Writing (unverified)·8 of 20 independently reproduced·$15/$75 per M tokens
Consensus: LiteLLM · Cross-check: models.dev, OpenRouter · See every model’s pricing →
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| Lech Mazur Writing | 84.7accuracy (%) | unverifiedT1 | 2025-08-05lechmazur/writing Github repository |
| SWE-Bench verified | 73.35accuracy (%) | reproduced· optimizedT1 | 2025-08-05Epoch AI |
| GPQA diamond | 69.7accuracy (%) | reproduced· optimizedT1 | 2025-08-05Epoch AI |
| OTIS Mock AIME 2024-2025 | 68.86accuracy (%) | reproduced· optimizedT1 | 2025-08-05Epoch AI |
| METR Time Horizons | 66.81accuracy (%) | unverifiedT1 | 2025-08-05METR - Measuring AI Ability to Complete Long Tasks |
| DTBench | 66.67accuracy (%) | unverifiedT1 | 2025-08-05https://conceptualreasoning.ai/dtbench |
| SimpleBench | 52accuracy (%) | unverifiedT1 | 2025-08-05SimpleBench Leaderboard |
| DeepResearch Bench | 48.3accuracy (%) | unverifiedT1 | 2025-08-05https://evals.futuresearch.ai/#drb |
| WeirdML | 45.86accuracy (%) | unverifiedT1 | 2025-08-05https://htihle.github.io/weirdml.html |
| LMCA | 43.62accuracy (%) | unverifiedT1 | 2025-08-05https://conceptualreasoning.ai/lmca |
| GDPval | 43.6accuracy (%) | unverifiedT2 | 2025-08-05 |
| Cybench | 42accuracy (%) | unverified· optimizedT1 | 2025-08-05Cybench leaderboard |
| Terminal Bench | 38accuracy (%) | unverified· optimizedT1 | 2025-08-05Terminal-Bench v2 Leaderboard |
| Mystery Game Puzzles | 12.98accuracy (%) | reproducedT1 | 2025-08-05Epoch AI |
| FrontierMath-Tiers-1-3-v2-Private | 12.63accuracy (%) | reproducedT1 | 2025-08-05Epoch AI |
| EBR-bench | 7.94accuracy (%) | reproducedT1 | 2025-08-05Epoch AI |
| HLE | 7.06accuracy (%) | unverifiedT2 | 2025-08-05 |
| VPCT | 2.5accuracy (%) | unverifiedT1 | 2025-08-05VPCT leaderboard |
| FrontierMath-Tier-4-v2-Private | 2.44accuracy (%) | reproducedT1 | 2025-08-05Epoch AI |
| Chess Puzzles | 2.15accuracy (%) | reproducedT1 | 2025-08-05Epoch AI |
Claims drawn from cited facts, not live model generation.
This model was developed by Anthropic. It was released in 2025, placing it in the most recent generation of models.
2 cited facts
This model is scored on 20 tracked benchmarks. Across those benchmarks, it holds no current top score. The results can be taken at face value, since the scores were independently reproduced by outside evaluation.
3 cited facts
On average this model trails the benchmark leader by 41.39 points and holds the top score on none of its benchmarks, so under the disclosed harness that wide gap reads as well behind the leaders, not at the frontier.
2 cited facts
6 facts cross-checked across data sources: 3 corroborated, 2 single-source, 1 disputed.
Source facts, citations, and refresh stamp for this record.
Sources: epoch_benchmarks, litellm_prices, modelsdev_models, openrouter_models
Claude Opus 4.1 is an AI model developed by Anthropic, released 2025-08-05. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
Claude Opus 4.1 has recorded scores on 20 benchmarks, each shown with its evidence status.
Claude Opus 4.1 has recorded scores on 20 benchmarks — Lech Mazur Writing, GPQA diamond, OTIS Mock AIME 2024-2025, DTBench, WeirdML, LMCA, and 14 more. The full table above shows each score with its evidence status.
8 of 20 recorded scores (40%) are independently reproduced rather than self-reported by the lab.
Claude Opus 4.1 has 20 tracked claims: 8 independently reproduced, 12 unverified.
Listed API pricing: $15 per million input tokens, $75 per million output tokens. See the pricing block for the full breakdown.