| Benchmark | GPT-4o | Kimi K3 |
|---|---|---|
| AA Agentic Index | 9.7 | 54.3 |
| AA Intelligence | 11.2 | 60 |
| AA-LCR | 55 | 82.7 |
| AA-Omniscience | -10.5 | 46 |
| arc_agi_1 | 4.5 | 65.7 |
| ARC-AGI-2 | 0 | 12.4 |
| Artificial Analysis Coding Index | 24.2 | 76.2 |
| critpt | 0 | 23.4 |
| gdpval | 0 | 59.4 |
| GPQA Diamond | 70.1 | 93.5 |
| HLE | 5.3 | 56 |
| mathvision | 30.4 | 97.8 |
| MMMU-Pro | 59.9 | 83.4 |
| OmniScience Accuracy | 23.7 | 47.6 |
| OmniScience Non-Hallucination | 62.1 | 49.1 |
| OSWorld-Verified | 5 | 84.8 |
| scicode | 33.4 | 58.7 |
| simplebench | 17.8 | 60.7 |
| Video-MME | 77.2 | 90 |
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.