| Benchmark | Claude 3.5 Sonnet | Qwen3.5 35B A3B |
|---|---|---|
| AA Intelligence | 10 | 29.9 |
| AI2D | 94.7 | 92.6 |
| Artificial Analysis Coding Index | 30.2 | 37 |
| C-Eval | 76.7 | 90.2 |
| GPQA Diamond | 67.2 | 84.5 |
| GSM8K | 96.9 | 90.1 |
| HLE | 3.9 | 47.4 |
| hmmt_feb_2025 | 1.7 | 89 |
| humaneval | 93.7 | 66.5 |
| ifeval | 90.1 | 91.9 |
| LiveCodeBench v6 | 37.2 | 74.6 |
| longbench_v2 | 41 | 59 |
| MathVista | 67.7 | 86.2 |
| mmlu | 88.3 | 81.1 |
| mmlu_redux | 88.9 | 93.3 |
| MMLU-Pro | 78 | 85.3 |
| MMMU | 72 | 81.4 |
| MMMU-Pro | 54.7 | 75.1 |
| MMStar | 62.2 | 81.9 |
| OCRBench | 790 | 91 |
| OpenBookQA | 97.2 | 44.2 |
| RealWorldQA | 60.1 | 84.1 |
| scicode | 36.6 | 37.7 |
| supergpqa | 48.2 | 63.4 |
| SWE-bench Verified | 50.8 | 70 |
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.