| Benchmark | Claude 3.5 Sonnet | GPT-5.1 |
|---|---|---|
| AA Intelligence | 10 | 37.5 |
| AIME 2025 | 7.4 | 94.2 |
| arena_vision | 1146 | 1250 |
| Artificial Analysis Coding Index | 30.2 | 49.4 |
| Fortress | 13 | 25.7 |
| frontiermath_tier_4 | 0 | 12.5 |
| GPQA Diamond | 67.2 | 88.1 |
| HLE | 3.9 | 28.5 |
| hmmt_feb_2025 | 1.7 | 96.3 |
| LiveCodeBench v6 | 37.2 | 87 |
| MMLU-Pro | 78 | 87 |
| MMMU | 72 | 85.4 |
| MMMU-Pro | 54.7 | 75.5 |
| scicode | 36.6 | 43.3 |
| SEAL VISTA | 38.7 | 43.8 |
| simplebench | 41.4 | 53.2 |
| SWE-bench Verified | 50.8 | 76.3 |
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.