| Benchmark | Claude 3.5 Sonnet | Qwen3.5 397B A17B |
|---|---|---|
| AA Intelligence | 10 | 34.3 |
| arena_vision | 1146 | 1249 |
| Artificial Analysis Coding Index | 30.2 | 48.2 |
| C-Eval | 76.7 | 93 |
| GPQA Diamond | 67.2 | 89.3 |
| GSM8K | 96.9 | 66.7 |
| HLE | 3.9 | 29 |
| hmmt_feb_2025 | 1.7 | 94.8 |
| ifeval | 90.1 | 92.6 |
| LiveCodeBench v6 | 37.2 | 83.6 |
| longbench_v2 | 41 | 63.2 |
| mmlu_redux | 88.9 | 94.9 |
| MMLU-Pro | 78 | 87.8 |
| MMMU | 72 | 85 |
| MMMU-Pro | 54.7 | 79 |
| MMStar | 62.2 | 83.8 |
| RealWorldQA | 60.1 | 83.9 |
| scicode | 36.6 | 42 |
| supergpqa | 48.2 | 70.4 |
| SWE-bench Verified | 50.8 | 76.4 |
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.