| Benchmark | Claude Opus 4.5 | Qwen2.5 Instruct 72B |
|---|---|---|
| AA Intelligence | 41.9 | 10 |
| AA-LCR | 76 | 20.3 |
| AA-Omniscience | 14 | -52.2 |
| Artificial Analysis Coding Index | 47.8 | 11.9 |
| C-Eval | 92.2 | 89.2 |
| critpt | 4.6 | 0 |
| Fortress | 13.6 | 56.4 |
| GPQA Diamond | 87 | 49.1 |
| HLE | 30.8 | 4.2 |
| IFBench | 58 | 36.9 |
| livebench | 76 | 52.3 |
| LiveCodeBench | 87 | 55.5 |
| longbench_v2 | 64.4 | 39.4 |
| mmlu_redux | 95.6 | 86.8 |
| MMLU-Pro | 90 | 71.6 |
| mmmlu | 90.8 | 74.8 |
| OmniScience Accuracy | 46.6 | 17.6 |
| OmniScience Non-Hallucination | 39 | 15.3 |
| scicode | 50 | 26.7 |
| SWE-bench Verified | 81.5 | 23.8 |
| Terminal-Bench Hard | 47 | 4.5 |
| τ²-Bench Telecom (AA run) | 89.5 | 34.5 |
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.