| Benchmark | DeepSeek-V3 | Kimi K3 |
|---|---|---|
| AA Agentic Index | 1.6 | 54.3 |
| AA Intelligence | 15.4 | 60 |
| AA-LCR | 41.3 | 82.7 |
| AA-Omniscience | -37.6 | 46 |
| Artificial Analysis Coding Index | 23 | 76.2 |
| critpt | 0 | 23.4 |
| gdpval | 0 | 59.4 |
| GPQA Diamond | 68.4 | 93.5 |
| HLE | 5.2 | 56 |
| OmniScience Accuracy | 25.4 | 47.6 |
| OmniScience Non-Hallucination | 23.3 | 49.1 |
| scicode | 35.8 | 58.7 |
| simplebench | 27.2 | 60.7 |
| TauBench V3 - Banking | 4.7 | 46 |
| Terminal-Bench 2.1 | 16.9 | 88.3 |
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.