| Benchmark | Kimi K3 | Qwen2.5 Instruct 72B |
|---|---|---|
| AA Intelligence | 60 | 10 |
| AA-LCR | 82.7 | 20.3 |
| AA-Omniscience | 46 | -52.2 |
| Artificial Analysis Coding Index | 76.2 | 11.9 |
| critpt | 23.4 | 0 |
| GPQA Diamond | 93.5 | 49.1 |
| HLE | 56 | 4.2 |
| OmniScience Accuracy | 47.6 | 17.6 |
| OmniScience Non-Hallucination | 49.1 | 15.3 |
| scicode | 58.7 | 26.7 |
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.