Across 50 shared benchmarks, Qwen3.5 122B A10B scores higher on 5 and Qwen3.6 Plus on 45. The widest gap is OmniScience Non-Hallucination, where Qwen3.6 Plus scores 68 against 12.9. Tracked API pricing per million tokens: Qwen3.5 122B A10B $0.40 in / $3.20 out, Qwen3.6 Plus $0.50 in / $3.00 out.
| Benchmark | Qwen3.5 122B A10B | Qwen3.6 Plus |
|---|---|---|
| AA Agentic Index | 21.3 | 29 |
| AA Intelligence | 32.8 | 40.5 |
| AA-LCR | 70.3 | 72.3 |
| AA-Omniscience | -41.5 | 2.6 |
| AI2D | 93.3 | 94.4 |
| Artificial Analysis Coding Index | 45.7 | 54.5 |
| C-Eval | 91.9 | 93.3 |
| CC-OCR | 81.8 | 83.4 |
| coding_arena_elo | 1358 | 1460 |
| critpt | 0.9 | 2.9 |
| DeepPlanning | 24.1 | 41.5 |
| DynaMath | 85.9 | 88 |
| ERQA | 62 | 65.7 |
| gdpval | 24.3 | 32 |
| GPQA Diamond | 86.6 | 90.4 |
| HLE | 47.5 | 28.8 |
| HMMT 2025 | 90.3 | 94.6 |
| IFBench | 76.1 | 75.2 |
| ifeval | 93.4 | 94.3 |
| include | 82.8 | 85.1 |
| LiveCodeBench v6 | 78.9 | 87.1 |
| longbench_v2 | 60.2 | 62 |
| mathvision | 86.2 | 88 |
| MLVU | 87.3 | 86.7 |
| mmlu_prox | 82.2 | 84.7 |
| mmlu_redux | 94 | 94.5 |
| MMLU-Pro | 86.7 | 88.5 |
| mmmlu | 86.7 | 89.5 |
| MMMU | 83.9 | 86 |
| MMMU-Pro | 76.9 | 78.8 |
| MMStar | 82.9 | 83.3 |
| OmniDocBench 1.5 | 89.8 | 91.2 |
| OmniScience Accuracy | 24.4 | 26.4 |
| OmniScience Non-Hallucination | 12.9 | 68 |
| OSWorld-Verified | 58 | 62.5 |
| RealWorldQA | 85.1 | 85.4 |
| scicode | 42 | 40.7 |
| screenspot_pro_no_tools | 70.4 | 68.2 |
| SimpleVQA | 0.6 | 0.7 |
| supergpqa | 67.1 | 71.6 |
| SWE-bench Verified | 72 | 78.8 |
| TauBench V3 - Banking | 15.3 | 20.8 |
| Terminal-Bench 2.0 | 49.4 | 61.6 |
| Terminal-Bench 2.1 | 47.6 | 61.4 |
| Terminal-Bench Hard | 31.1 | 43.9 |
| VideoMMMU | 82 | 84 |
| VITA-Bench | 33.6 | 44.3 |
| WideSearch | 60.5 | 74.3 |
| τ²-Bench Telecom (AA run) | 93.6 | 97.7 |
| τ³-Bench | 13.6 | 70.7 |
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, alibaba-official), otherwise the lowest tracked offer.