| Benchmark | Kimi K3 | Qwen3.6 27B |
|---|---|---|
| AA Agentic Index | 54.3 | 27.5 |
| AA Intelligence | 60 | 37.7 |
| AA-LCR | 82.7 | 73.3 |
| AA-Omniscience | 46 | -20 |
| Agents' Last Exam | 28.3 | 27.3 |
| aime_2026 | 96.7 | 94.1 |
| Artificial Analysis Coding Index | 76.2 | 53.7 |
| baby_vision | 85.7 | 28.9 |
| charxiv_rq | 84.8 | 78.4 |
| critpt | 23.4 | 1.1 |
| DeepSWE 1.1 | 69 | 13.3 |
| gdpval | 59.4 | 32 |
| GPQA Diamond | 93.5 | 87.8 |
| HLE | 56 | 24 |
| hmmt_feb_2026 | 97 | 84.3 |
| JobBench | 54.3 | 21.8 |
| livebench_agentic_coding | 62.2 | 39.3 |
| livebench_coding | 81.5 | 71.8 |
| livebench_data_analysis | 78.7 | 70.4 |
| livebench_instruction_following | 71.4 | 53.2 |
| livebench_language | 85.5 | 63.3 |
| livebench_math | 84.4 | 79.9 |
| livebench_reasoning | 90.7 | 70.3 |
| mathvision | 97.8 | 85.1 |
| MMMU-Pro | 83.4 | 75.8 |
| OmniScience Accuracy | 47.6 | 19.6 |
| OmniScience Non-Hallucination | 49.1 | 50.7 |
| OSWorld-Verified | 84.8 | 63.9 |
| scicode | 58.7 | 39.8 |
| TauBench V3 - Banking | 46 | 16.7 |
| Terminal-Bench 2.1 | 88.3 | 60.7 |
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.