| Benchmark | Kimi K2 (Reasoning) | Kimi K3 |
|---|---|---|
| AA Agentic Index | 47.9 | 54.3 |
| AA Intelligence | 33.5 | 60 |
| AA-LCR | 70.3 | 82.7 |
| AA-Omniscience | -21.4 | 46 |
| Artificial Analysis Coding Index | 34.8 | 76.2 |
| browsecomp | 60.2 | 91.2 |
| critpt | 2.6 | 23.4 |
| gdpval | 24.5 | 59.4 |
| GPQA Diamond | 84.5 | 93.5 |
| HLE | 23.9 | 56 |
| OmniScience Accuracy | 30.9 | 47.6 |
| OmniScience Non-Hallucination | 25.8 | 49.1 |
| ResearchRubrics | 56.2 | 76.2 |
| scicode | 44.8 | 58.7 |
| simplebench | 39.6 | 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.