| Benchmark | Claude 3.5 Sonnet | Kimi K2.5 |
|---|---|---|
| AA Intelligence | 10 | 36 |
| AIME 2025 | 7.4 | 96.1 |
| arena_vision | 1146 | 1247 |
| Artificial Analysis Coding Index | 30.2 | 46.8 |
| Fortress | 13 | 41.1 |
| frontiermath_tier_4 | 0 | 4.2 |
| GPQA Diamond | 67.2 | 87.9 |
| HLE | 3.9 | 50.2 |
| hmmt_feb_2025 | 1.7 | 95.4 |
| LiveCodeBench v6 | 37.2 | 85 |
| longbench_v2 | 41 | 61 |
| MathVista | 67.7 | 90.1 |
| MMLU-Pro | 78 | 87.1 |
| MMMU-Pro | 54.7 | 78.5 |
| OCRBench | 790 | 92.3 |
| scicode | 36.6 | 49 |
| SEAL VISTA | 38.7 | 41.9 |
| simplebench | 41.4 | 46.8 |
| SWE-bench Verified | 50.8 | 76.8 |
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.