| Benchmark | Claude 4 Opus | Qwen3.6 Plus |
|---|---|---|
| AA Intelligence | 31.7 | 40.5 |
| AA-LCR | 40 | 72.3 |
| Artificial Analysis Coding Index | 34 | 54.5 |
| frontiermath_tier_4 | 4.2 | 8.3 |
| GPQA Diamond | 79.6 | 90.4 |
| HLE | 12.3 | 28.8 |
| HMMT 2025 | 15.9 | 94.6 |
| IFBench | 53.7 | 75.2 |
| ifeval | 87.4 | 94.3 |
| livebench | 74.6 | 70.8 |
| LiveCodeBench v6 | 47.4 | 87.1 |
| longbench_v2 | 55.6 | 62 |
| mmlu_redux | 94.2 | 94.5 |
| MMLU-Pro | 86.6 | 88.5 |
| mmmlu | 88.8 | 89.5 |
| MMMU | 73.7 | 86 |
| scicode | 40.9 | 40.7 |
| supergpqa | 56.5 | 71.6 |
| SWE-bench Verified | 79.4 | 78.8 |
| Terminal-Bench Hard | 31.1 | 43.9 |
| τ²-Bench Telecom (AA run) | 73.4 | 97.7 |
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.