| Benchmark | Claude 3.7 Sonnet | Kimi K3 |
|---|---|---|
| AA Agentic Index | 37 | 54.3 |
| AA Intelligence | 27.6 | 60 |
| AA-LCR | 62.3 | 82.7 |
| AA-Omniscience | -0.7 | 46 |
| arc_agi_1 | 13.6 | 65.7 |
| ARC-AGI-2 | 0 | 12.4 |
| Artificial Analysis Coding Index | 36.4 | 76.2 |
| critpt | 0.9 | 23.4 |
| gdpval | 27.4 | 59.4 |
| GPQA Diamond | 84.8 | 93.5 |
| HLE | 9.7 | 56 |
| MMMU-Pro | 60.1 | 83.4 |
| OmniScience Accuracy | 28.2 | 47.6 |
| OmniScience Non-Hallucination | 60 | 49.1 |
| scicode | 40.3 | 58.7 |
| simplebench | 46.4 | 60.7 |
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.