| Benchmark | Opus 5 | Qwen3.5 397B A17B |
|---|---|---|
| AA Agentic Index | 59.2 | 53.3 |
| AA Intelligence | 63.1 | 34.3 |
| AA-LCR | 78.7 | 72.7 |
| AA-Omniscience | 37.1 | -30.8 |
| arena_elo | 1487 | 1442 |
| arena_text_factuality | 1481 | 1445 |
| Artificial Analysis Coding Index | 78 | 48.2 |
| browsecomp | 90.8 | 69 |
| coding_arena_elo | 1662 | 1399 |
| critpt | 29.1 | 1.7 |
| gdpval | 67.2 | 35.8 |
| GDPval-AA v2 | 1861 | 962 |
| GPQA Diamond | 93.7 | 89.3 |
| HLE | 64.7 | 29 |
| HLE (with tools) | 64.7 | 48.3 |
| MMMU-Pro | 84.7 | 79 |
| OmniScience Accuracy | 60.9 | 30.8 |
| OmniScience Non-Hallucination | 40.5 | 17.3 |
| scicode | 55.7 | 42 |
| simpleqa_verified | 56.7 | 26 |
| SWE-bench Multilingual | 89.5 | 69.3 |
| SWE-bench Pro | 79.2 | 50.9 |
| SWE-bench Verified | 96 | 76.4 |
| TauBench V3 - Banking | 44.7 | 13.4 |
| Terminal-Bench 2.1 | 89.1 | 51.3 |
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.