| Benchmark | Claude 4 Opus | Qwen3.5 397B A17B |
|---|---|---|
| AA Intelligence | 31.7 | 34.3 |
| AA-LCR | 40 | 72.7 |
| arena_vision | 1207 | 1249 |
| Artificial Analysis Coding Index | 34 | 48.2 |
| GPQA Diamond | 79.6 | 89.3 |
| HLE | 12.3 | 29 |
| HMMT 2025 | 15.9 | 92.7 |
| hmmt_feb_2025 | 60 | 94.8 |
| IFBench | 53.7 | 78.8 |
| ifeval | 87.4 | 92.6 |
| LiveCodeBench v6 | 47.4 | 83.6 |
| longbench_v2 | 55.6 | 63.2 |
| mmlu_redux | 94.2 | 94.9 |
| MMLU-Pro | 86.6 | 87.8 |
| mmmlu | 88.8 | 88.5 |
| MMMU | 73.7 | 85 |
| multichallenge | 58.6 | 67.6 |
| scicode | 40.9 | 42 |
| supergpqa | 56.5 | 70.4 |
| SWE-bench Verified | 79.4 | 76.4 |
| Terminal-Bench Hard | 31.1 | 40.9 |
| τ²-Bench Telecom (AA run) | 73.4 | 95.6 |
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.