Across 33 shared benchmarks, GPT-5.1 scores higher on 26 and Qwen3.5 122B A10B on 7. The widest gap is critpt, where GPT-5.1 scores 4.9 against 0.9. Qwen3.5 122B A10B is the cheaper of the two on tracked API pricing ($0.40 against $1.25 per million input tokens).
| Benchmark | GPT-5.1 | Qwen3.5 122B A10B |
|---|---|---|
| AA Agentic Index | 32.2 | 21.3 |
| AA Intelligence | 37.5 | 32.8 |
| AA-LCR | 76.7 | 70.3 |
| AA-Omniscience | 5.4 | -41.5 |
| arena_vision | 1250 | 1246 |
| Artificial Analysis Coding Index | 49.4 | 45.7 |
| browsecomp | 50.8 | 63.8 |
| coding_arena_elo | 1421 | 1358 |
| critpt | 4.9 | 0.9 |
| gdpval | 25 | 24.3 |
| GPQA Diamond | 88.1 | 86.6 |
| HLE | 28.5 | 47.5 |
| IFBench | 72.9 | 76.1 |
| LiveCodeBench v6 | 87 | 78.9 |
| MMLU-Pro | 87 | 86.7 |
| mmmlu | 91 | 86.7 |
| MMMU | 85.4 | 83.9 |
| MMMU-Pro | 75.5 | 76.9 |
| multichallenge | 63.4 | 61.5 |
| OmniScience Accuracy | 37.7 | 24.4 |
| OmniScience Non-Hallucination | 48.1 | 12.9 |
| scicode | 43.3 | 42 |
| SWE-bench Verified | 76.3 | 72 |
| TauBench V3 - Banking | 15.9 | 15.3 |
| Terminal-Bench 2.0 | 47.6 | 49.4 |
| Terminal-Bench 2.1 | 52.4 | 47.6 |
| Terminal-Bench Hard | 45.5 | 31.1 |
| vectara_answer_rate | 100 | 99.8 |
| vectara_avg_summary_length | 254.4 | 86.4 |
| vectara_factual_consistency | 89.1 | 88.8 |
| vectara_hallucination_rate ↓ | 12.1 | 11.2 |
| τ²-Bench Telecom (AA run) | 81.9 | 93.6 |
| τ³-Bench | 14 | 13.6 |
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 (direct, direct), otherwise the lowest tracked offer.