| Benchmark | Kimi K3 | o4-mini |
|---|---|---|
| AA Agentic Index | 54.3 | 36.1 |
| AA Intelligence | 60 | 26.1 |
| AA-LCR | 82.7 | 60 |
| AA-Omniscience | 46 | -35.7 |
| arc_agi_1 | 65.7 | 58.7 |
| ARC-AGI-2 | 12.4 | 6.1 |
| Artificial Analysis Coding Index | 76.2 | 25.6 |
| browsecomp | 91.2 | 51.5 |
| critpt | 23.4 | 0.6 |
| gdpval | 59.4 | 25.4 |
| GPQA Diamond | 93.5 | 81.4 |
| HLE | 56 | 16.5 |
| MMMU-Pro | 83.4 | 69.2 |
| OmniScience Accuracy | 47.6 | 24.8 |
| OmniScience Non-Hallucination | 49.1 | 19.5 |
| scicode | 58.7 | 46.5 |
| simplebench | 60.7 | 38.7 |
| simpleqa_verified | 42.7 | 23.9 |
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.