| Benchmark | Claude 3.5 Sonnet | GPT-5.4 |
|---|---|---|
| AA Intelligence | 10 | 53.1 |
| arena_vision | 1146 | 1277 |
| Artificial Analysis Coding Index | 30.2 | 71.1 |
| Chinese SimpleQA (C-SimpleQA) | 56.8 | 76.8 |
| Codeforces | 717 | 3168 |
| frontiermath_tier_4 | 0 | 27.1 |
| GPQA Diamond | 67.2 | 93 |
| HLE | 3.9 | 43.7 |
| MMLU-Pro | 78 | 87.5 |
| MMMU-Pro | 54.7 | 81.2 |
| scicode | 36.6 | 56.6 |
| SEAL VISTA | 38.7 | 50.9 |
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.