| Benchmark | Claude 3.5 Sonnet | Qwen3.6 35B A3B |
|---|---|---|
| AA Intelligence | 10 | 32.1 |
| AI2D | 94.7 | 92.7 |
| Artificial Analysis Coding Index | 30.2 | 41.9 |
| C-Eval | 76.7 | 90 |
| GPQA Diamond | 67.2 | 86 |
| HLE | 3.9 | 22.2 |
| hmmt_feb_2025 | 1.7 | 90.7 |
| LiveCodeBench v6 | 37.2 | 80.4 |
| MathVista | 67.7 | 86.4 |
| mmlu_redux | 88.9 | 93.3 |
| MMLU-Pro | 78 | 85.6 |
| MMMU | 72 | 81.7 |
| MMMU-Pro | 54.7 | 75.3 |
| RealWorldQA | 60.1 | 85.3 |
| scicode | 36.6 | 35.8 |
| supergpqa | 48.2 | 64.7 |
| SWE-bench Verified | 50.8 | 73.4 |
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.