Across 42 shared benchmarks, GPT-5.2 scores higher on 32 and Qwen3.5 122B A10B on 10. The widest gap is OmniScience Non-Hallucination, where GPT-5.2 scores 38.4 against 12.9. Qwen3.5 122B A10B is the cheaper of the two on tracked API pricing ($0.40 against $1.75 per million input tokens).
| Benchmark | GPT-5.2 | Qwen3.5 122B A10B |
|---|---|---|
| AA Agentic Index | 60.2 | 21.3 |
| AA Intelligence | 43.3 | 32.8 |
| AA-LCR | 79.3 | 70.3 |
| AA-Omniscience | 0.3 | -41.5 |
| arena_vision | 1245 | 1246 |
| Artificial Analysis Coding Index | 48.7 | 45.7 |
| browsecomp | 65.8 | 63.8 |
| browsecomp_zh | 76.1 | 69.9 |
| coding_arena_elo | 1418 | 1358 |
| critpt | 11.6 | 0.9 |
| gdpval | 48.3 | 24.3 |
| GPQA Diamond | 92.4 | 86.6 |
| HLE | 45.5 | 47.5 |
| HMMT 2025 | 99.4 | 90.3 |
| IFBench | 75.4 | 76.1 |
| longbench_v2 | 54.5 | 60.2 |
| mathvision | 83 | 86.2 |
| MathVista | 82.8 | 87.4 |
| MMLU-Pro | 87 | 86.7 |
| mmmlu | 89.6 | 86.7 |
| MMMU-Pro | 79.5 | 76.9 |
| MMVU | 80.8 | 74.7 |
| OCRBench | 80.7 | 92.1 |
| OmniDocBench 1.5 | 85.7 | 89.8 |
| OmniScience Accuracy | 44.3 | 24.4 |
| OmniScience Non-Hallucination | 38.4 | 12.9 |
| scicode | 52.1 | 42 |
| screenspot_pro_no_tools | 86.3 | 70.4 |
| Seal-0 | 45 | 44.1 |
| SimpleVQA | 55.8 | 0.6 |
| SWE-bench Verified | 80 | 72 |
| TauBench V3 - Banking | 11.1 | 15.3 |
| Terminal-Bench 2.0 | 54 | 49.4 |
| Terminal-Bench Hard | 62.2 | 31.1 |
| vectara_answer_rate | 100 | 99.8 |
| vectara_avg_summary_length | 186.3 | 86.4 |
| vectara_factual_consistency | 89.2 | 88.8 |
| vectara_hallucination_rate ↓ | 10.8 | 11.2 |
| VideoMMMU | 85.9 | 82 |
| ZeroBench | 9 | 0.1 |
| τ²-Bench Telecom (AA run) | 98.7 | 93.6 |
| τ³-Bench | 12 | 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.