| Benchmark | Claude 3.7 Sonnet | GPT-5.1 |
|---|---|---|
| AA Agentic Index | 37 | 32.2 |
| AA Intelligence | 27.6 | 37.5 |
| AA-LCR | 62.3 | 76.7 |
| AA-Omniscience | -0.7 | 5.4 |
| AIME 2025 | 54.8 | 94.2 |
| arc_agi_1 | 13.6 | 72.8 |
| ARC-AGI-2 | 0 | 17.6 |
| arena_vision | 1195 | 1250 |
| Artificial Analysis Coding Index | 36.4 | 49.4 |
| critpt | 0.9 | 4.9 |
| Fortress | 38 | 25.7 |
| gdpval | 27.4 | 25 |
| GPQA Diamond | 84.8 | 88.1 |
| HLE | 9.7 | 28.5 |
| hmmt_feb_2025 | 31.7 | 96.3 |
| IFBench | 48.3 | 72.9 |
| mmmlu | 86.1 | 91 |
| MMMU | 75 | 85.4 |
| MMMU-Pro | 60.1 | 75.5 |
| multichallenge | 51.6 | 63.4 |
| MultiNRC | 27.8 | 49 |
| OmniScience Accuracy | 28.2 | 37.7 |
| OmniScience Non-Hallucination | 60 | 48.1 |
| scicode | 40.3 | 43.3 |
| SEAL VISTA | 48.2 | 43.8 |
| simplebench | 46.4 | 53.2 |
| swe_bench_bash | 52.8 | 66 |
| SWE-bench Verified | 70.3 | 76.3 |
| Terminal-Bench Hard | 21.2 | 45.5 |
| τ²-Bench Telecom (AA run) | 54.7 | 81.9 |
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.