Across 20 shared benchmarks, Opus 5 scores higher on 19 and Qwen3.5 122B A10B on 1. The widest gap is OmniScience Non-Hallucination, where Opus 5 scores 40.5 against 12.9. Qwen3.5 122B A10B is the cheaper of the two on tracked API pricing ($0.40 against $5.00 per million input tokens).
| Benchmark | Opus 5 | Qwen3.5 122B A10B |
|---|---|---|
| AA Agentic Index | 59.2 | 21.3 |
| AA Intelligence | 63.1 | 32.8 |
| AA-LCR | 78.7 | 70.3 |
| AA-Omniscience | 37.1 | -41.5 |
| arena_vision | 1323 | 1246 |
| Artificial Analysis Coding Index | 78 | 45.7 |
| browsecomp | 90.8 | 63.8 |
| coding_arena_elo | 1663 | 1358 |
| critpt | 29.1 | 0.9 |
| gdpval | 66.7 | 24.3 |
| GPQA Diamond | 93.7 | 86.6 |
| HLE | 64.7 | 47.5 |
| MMMU-Pro | 84.7 | 76.9 |
| OmniScience Accuracy | 60.9 | 24.4 |
| OmniScience Non-Hallucination | 40.5 | 12.9 |
| OSWorld-Verified | 4 | 58 |
| scicode | 55.7 | 42 |
| SWE-bench Verified | 96 | 72 |
| TauBench V3 - Banking | 44.7 | 15.3 |
| Terminal-Bench 2.1 | 89.1 | 47.6 |
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. Quoted rates are the price-setter row we currently track for each model — its direct or vendor-official listing where one exists (anthropic-official, direct), otherwise the lowest tracked offer.