| Benchmark | Claude Sonnet 4.5 | Gemini 1.5 Pro |
|---|---|---|
| AA Intelligence | 37.4 | 10 |
| AI2D | 87 | 80.9 |
| AIR-Bench 2024 | 89.8 | 82.8 |
| ARC-AGI-2 | 13.6 | 0.8 |
| Artificial Analysis Coding Index | 52.1 | 23.6 |
| Fortress | 17 | 53.9 |
| GPQA Diamond | 83.4 | 59.1 |
| HLE | 19.8 | 4.9 |
| LiveCodeBench | 71 | 30.5 |
| MathVista | 79.8 | 70.6 |
| MMLU-Pro | 88.2 | 75.8 |
| MMMU | 79.6 | 68.4 |
| MMMU (val) (Pass@1) | 77.8 | 54.1 |
| MMMU-Pro | 68.7 | 55 |
| mrcr | 55.4 | 82.6 |
| scicode | 45 | 29.5 |
| SEAL VISTA | 48.8 | 37.1 |
| simplebench | 54.3 | 27.1 |
| SWE-bench Verified | 82 | 34.2 |
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.