| Benchmark | Claude 3.5 Sonnet | Qwen3.6 27B |
|---|---|---|
| AA Intelligence | 10 | 37.7 |
| Artificial Analysis Coding Index | 30.2 | 53.7 |
| C-Eval | 76.7 | 91.4 |
| GPQA Diamond | 67.2 | 87.8 |
| HLE | 3.9 | 24 |
| hmmt_feb_2025 | 1.7 | 93.8 |
| humaneval | 93.7 | 78.2 |
| LiveCodeBench v6 | 37.2 | 83.9 |
| MathVista | 67.7 | 87.4 |
| mmlu_redux | 88.9 | 93.5 |
| MMLU-Pro | 78 | 86.2 |
| MMMU | 72 | 82.9 |
| MMMU-Pro | 54.7 | 75.8 |
| MMStar | 62.2 | 81.4 |
| OCRBench | 790 | 89.4 |
| RealWorldQA | 60.1 | 84.1 |
| scicode | 36.6 | 39.8 |
| simpleqa | 28.4 | 95.7 |
| supergpqa | 48.2 | 66 |
| SWE-bench Verified | 50.8 | 77.2 |
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.