| Benchmark | Kimi K3 | Qwen3.6 35B A3B |
|---|---|---|
| AA Agentic Index | 54.3 | 21.6 |
| AA Intelligence | 60 | 32.1 |
| AA-LCR | 82.7 | 66.7 |
| AA-Omniscience | 46 | -22.2 |
| aime_2026 | 96.7 | 92.7 |
| Artificial Analysis Coding Index | 76.2 | 41.9 |
| browsecomp | 91.2 | 48.7 |
| charxiv_rq | 84.8 | 78 |
| critpt | 23.4 | 0.3 |
| gdpval | 59.4 | 27.8 |
| GDPval-AA v2 | 1687 | 1015 |
| GPQA Diamond | 93.5 | 86 |
| HLE | 56 | 22.2 |
| hmmt_feb_2026 | 97 | 83.6 |
| MCP Atlas | 84.2 | 62.8 |
| MMMU-Pro | 83.4 | 75.3 |
| OmniScience Accuracy | 47.6 | 18.8 |
| OmniScience Non-Hallucination | 49.1 | 49.5 |
| scicode | 58.7 | 35.8 |
| TauBench V3 - Banking | 46 | 9.3 |
| Terminal-Bench 2.1 | 88.3 | 44.9 |
| toolathlon | 73.2 | 26.9 |
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.