| Benchmark | Claude Sonnet 4 | o1 |
|---|---|---|
| AA Agentic Index | 39.2 | 31.1 |
| AA Intelligence | 29.8 | 24 |
| AA-LCR | 69.7 | 63.3 |
| AA-Omniscience | 0.2 | -10.6 |
| aider_polyglot | 70.7 | 61.7 |
| AIME 2024 | 43.4 | 79.2 |
| AIME 2025 | 74 | 81.7 |
| AIR-Bench 2024 | 88.3 | 80 |
| anthropic_red_team | 99.2 | 98.3 |
| arena_vision | 1207 | 1193 |
| Artificial Analysis Coding Index | 37.6 | 39.7 |
| bbq | 97.9 | 97.3 |
| critpt | 1.1 | 0.3 |
| Fortress | 24.4 | 19.4 |
| gdpval | 31.2 | 11.5 |
| GPQA Diamond | 78 | 75.7 |
| harmbench | 98 | 96.3 |
| HLE | 10.7 | 7.7 |
| IFBench | 55 | 70.3 |
| livebench | 74.8 | 52.3 |
| math_500_em | 94 | 96.4 |
| mmlu | 91.5 | 91.8 |
| mmmlu | 86.5 | 87.7 |
| MMMU | 74.4 | 77.6 |
| OmniScience Accuracy | 22.7 | 34.7 |
| OmniScience Non-Hallucination | 70.9 | 30.7 |
| scicode | 40 | 35.8 |
| SEAL VISTA | 45.5 | 45.3 |
| simple_safety_tests | 100 | 99 |
| simplebench | 45.5 | 41.7 |
| simpleqa | 15.9 | 47 |
| SWE-bench Verified | 80.2 | 48.9 |
| TAU-bench (airline) | 60 | 50 |
| TAU-bench (retail) | 80.5 | 70.8 |
| Terminal-Bench Hard | 31.1 | 12.9 |
| xstest | 97.2 | 97 |
| τ²-Bench Telecom (AA run) | 64.6 | 62.6 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.