Across 26 shared benchmarks, Claude 4 Opus scores higher on 4 and Qwen3.5 122B A10B on 21, with 1 level. The widest gap is HMMT 2025, where Qwen3.5 122B A10B scores 90.3 against 15.9. Qwen3.5 122B A10B is the cheaper of the two on tracked API pricing ($0.40 against $15.00 per million input tokens).
| Benchmark | Claude 4 Opus | Qwen3.5 122B A10B |
|---|---|---|
| AA Intelligence | 31.7 | 32.8 |
| AA-LCR | 40 | 70.3 |
| arena_vision | 1207 | 1246 |
| Artificial Analysis Coding Index | 34 | 45.7 |
| GPQA Diamond | 79.6 | 86.6 |
| HLE | 12.3 | 47.5 |
| HMMT 2025 | 15.9 | 90.3 |
| IFBench | 53.7 | 76.1 |
| ifeval | 87.4 | 93.4 |
| LiveCodeBench v6 | 47.4 | 78.9 |
| longbench_v2 | 55.6 | 60.2 |
| mmlu_redux | 94.2 | 94 |
| MMLU-Pro | 86.6 | 86.7 |
| mmmlu | 88.8 | 86.7 |
| MMMU | 73.7 | 83.9 |
| multichallenge | 58.6 | 61.5 |
| OJBench | 19.6 | 39.5 |
| scicode | 40.9 | 42 |
| supergpqa | 56.5 | 67.1 |
| SWE-bench Verified | 79.4 | 72 |
| Terminal-Bench Hard | 31.1 | 31.1 |
| vectara_answer_rate | 91 | 99.8 |
| vectara_avg_summary_length | 123.2 | 86.4 |
| vectara_factual_consistency | 88 | 88.8 |
| vectara_hallucination_rate ↓ | 12 | 11.2 |
| τ²-Bench Telecom (AA run) | 73.4 | 93.6 |
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.