← Models
Qwen3.5-122B-A10B
Alibaba Cloud / Qwen Team · open weight · qwen3.5-122b-a10b
Qwen3.5-122B-A10B is a multimodal Mixture-of-Experts model with 122 billion total parameters and 10 billion activated parameters. It combines strong reasoning, coding, long-context, and visual understanding performance with production-friendly efficiency and a native 262K context window.
Benchmark scores
| Benchmark | Score |
|---|---|
| CountBench | 0.97 |
| VLMsAreBlind | 0.97 |
| MMLU-Redux | 0.94 |
| IFEval | 0.93 |
| AI2D | 0.93 |
| V* | 0.93 |
| MMBench-V1.1 | 0.93 |
| OCRBench | 0.92 |
| HMMT 2025 | 0.91 |
| RefCOCO-avg | 0.91 |
| HMMT25 | 0.90 |
| OmniDocBench 1.5 | 0.90 |
| MathVista-Mini | 0.87 |
| MLVU | 0.87 |
| VideoMME w sub. | 0.87 |
| MMLU-Pro | 0.87 |
| GPQA | 0.87 |
| MathVision | 0.86 |
| DynaMath | 0.86 |
| RealWorldQA | 0.85 |
| MMMU | 0.84 |
| t2-bench | 0.80 |
| MMMU-Pro | 0.77 |
| BFCL-V4 | 0.72 |
| SWE-Bench Verified | 0.72 |
| Humanity's Last Exam | 0.47 |
Pricing
- DeepInfra$0.29 / $2.40
Input / output per 1M tokens
AA metrics
No Artificial Analysis link yet.
Arena Elo
- No Arena snapshot linked.