Anthropic·released 2026-06-091 source
Claude Fable 5 benchmark scores: 25 benchmarks tracked, leading 5. 24% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 5 of 25 benchmarks·best result 100 on OTIS Mock AIME 2024-2025 (reproduced)·6 of 25 independently reproduced·$10/$50 per M tokens
Head-to-headClaude Fable 5 vs Claude Opus 5.5Consensus: LiteLLM · Cross-check: models.dev, OpenRouter · See every model’s pricing →
Leads on: ARC-AGI, OTIS Mock AIME 2024-2025, PostTrainBench, SimpleBench, Surface Evolver Bench
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| OTIS Mock AIME 2024-2025 | 100accuracy (%) | reproduced· optimizedT1 | 2026-06-09Epoch AI |
| ARC-AGI | 98.5accuracy (%) | unverified· optimizedT1 | 2026-06-09https://arcprize.org/leaderboard |
| DTBench | 97.33accuracy (%) | unverifiedT1 | 2026-06-09https://conceptualreasoning.ai/dtbench |
| Surface Evolver Bench | 95accuracy (%) | unverifiedT1 | 2026-06-09https://yhenon.github.io/surface-evolver-llm-eval/ |
| ProofBench | 95accuracy (%) | unverifiedT1 | 2026-06-09https://www.vals.ai/benchmarks/proof_bench |
| WeirdML | 91.94accuracy (%) | unverifiedT1 | 2026-06-09https://htihle.github.io/weirdml.html |
| FrontierMath-Tier-4-v2-Private | 90.2accuracy (%) | unverifiedT2 | 2026-06-09 |
| ARC-AGI-2 | 89.17accuracy (%) | unverifiedT2 | 2026-06-09 |
| FrontierMath-Tiers-1-3-v2-Private | 87.02accuracy (%) | reproducedT1 | 2026-06-09Epoch AI |
| GPQA diamond | 81.14accuracy (%) | reproduced· optimizedT1 | 2026-06-09Epoch AI |
| GSO-Bench | 78.43accuracy (%) | unverifiedT1 | 2026-06-09https://gso-bench.github.io/index.html |
| SimpleBench | 78.28accuracy (%) | unverifiedT1 | 2026-06-09SimpleBench Leaderboard |
| LMCA | 71.94accuracy (%) | unverifiedT1 | 2026-06-09https://conceptualreasoning.ai/lmca |
| SimpleQA Verified | 70.7accuracy (%) | unverifiedT2 | 2026-06-09 |
| DeepSWE | 69.91accuracy (%) | unverifiedT1 | 2026-06-09https://deepswe.datacurve.ai/ |
| MirrorCode | 63.89accuracy (%) | reproducedT1 | 2026-06-09Epoch AI |
| APEX-Agents | 63.6accuracy (%) | unverifiedT2 | 2026-06-09 |
| FrontierCode | 53.48accuracy (%) | unverifiedT1 | 2026-06-09https://cognition.com/frontiercode |
| Mystery Game Puzzles | 47.12accuracy (%) | reproducedT1 | 2026-06-09Epoch AI |
| FrontierSWE | 46.96accuracy (%) | unverifiedT1 | 2026-06-09https://www.frontierswe.com/ |
| Chess Puzzles | 44.23accuracy (%) | unverifiedT2 | 2026-06-09 |
| PostTrainBench | 41.79accuracy (%) | unverifiedT2 | 2026-06-09 |
| EBR-bench | 39.52accuracy (%) | reproducedT1 | 2026-06-09Epoch AI |
| Remote Labor Index | 16.1accuracy (%) | unverifiedT2 | 2026-06-09 |
| Furniture Assembly | 8.33accuracy (%) | unverifiedT2 | 2026-06-09 |
Claims drawn from cited facts, not live model generation.
This model was developed by Anthropic and released in 2026.
2 cited facts
This model has scores on 25 tracked benchmarks and currently leads 5 of them. The record’s evidence status is independently reproduced.
3 cited facts
Its mean gap to the best recorded score across its tracked benchmarks is 11.88 points.
1 cited fact
6 facts cross-checked across data sources: 3 corroborated, 2 single-source, 1 disputed.
Source facts, citations, and refresh stamp for this record.
Sources: curated_capability_claims, epoch_benchmarks, litellm_prices, modelsdev_models +1 more
Claude Fable 5 is an AI model developed by Anthropic, released 2026-06-09. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
Claude Fable 5 has recorded scores on 25 benchmarks, each shown with its evidence status.
Claude Fable 5 has recorded scores on 25 benchmarks — GPQA diamond, OTIS Mock AIME 2024-2025, DTBench, WeirdML, LMCA, SimpleQA Verified, and 19 more. The full table above shows each score with its evidence status.
6 of 25 recorded scores (24%) are independently reproduced rather than self-reported by the lab.
Claude Fable 5 has 25 tracked claims: 6 independently reproduced, 19 unverified.
Listed API pricing: $10 per million input tokens, $50 per million output tokens. See the pricing block for the full breakdown.