Alibaba Cloud / Qwen Team model product
Qwen3 VL 235B A22B Thinking
Qwen3-VL-235B-A22B-Thinking is the most powerful vision-language model in the Qwen series, featuring 236B parameters with MoE architecture for reasoning-enhanced multimodal understanding.
Updated Aug 12, 2026. Default version: Qwen3 VL 235B A22B Thinking
LLMBoard score45.8Qwen3 VL 235B A22B Thinking
Coverage100%58 benchmark families
Context window262.1KTokens
Official input priceN/AOfficial price unavailable
Qwen3 VL 235B A22B Thinking Specifications
Technical details for the model's default version.
- Version
- Qwen3 VL 235B A22B Thinking
- Released
- Sep 22, 2025
- Knowledge cutoff
- Unknown
- Parameters
- 236B
- Context window
- 262.1K
- Max output
- 262.1K
- Inputs
- image, text, video
- Outputs
- text
- Open weights
- No
- License
- Apache 2.0
Qwen3 VL 235B A22B Thinking Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Qwen3 VL 235B A22B Thinking Benchmark Results
Benchmark scores for Qwen3 VL 235B A22B Thinking.
| ARKitScenes | 0.537 points | 01 | 1 | 100.0% | C | |
| LiveBench 20241125 | 79.6% | 01 | 14 | 100.0% | C | |
| MathVerse-Mini | 0.85 points | 01 | 1 | 100.0% | C | |
| MIABench | 0.927 points | 01 | 1 | 100.0% | C | |
| MMMUval | 80.6% | 01 | 4 | 100.0% | C | |
| Objectron | 0.712 points | 01 | 1 | 100.0% | C | |
| OCRBench-V2 (zh) | 63.5% | 01 | 11 | 100.0% | C | |
| OSWorld-G | 0.683 points | 01 | 1 | 100.0% | C | |
| RoboSpatialHome | 0.739 points | 01 | 1 | 100.0% | C | |
| SIFO | 0.773 points | 01 | 1 | 100.0% | C | |
| SIFO-Multiturn | 0.711 points | 01 | 1 | 100.0% | C | |
| ZebraLogic | 97.3% | 01 | 8 | 100.0% | C | |
| CharadesSTA | 63.5% | 02 | 12 | 90.9% | C | |
| Design2Code | 0.934 points | 02 | 2 | 0.0% | C | |
| InfoVQAtest | 89.5% | 02 | 12 | 90.9% | C | |
| MuirBench | 80.1% | 02 | 11 | 90.0% | C | |
| RefSpatialBench | 0.699 points | 02 | 6 | 80.0% | C | |
| EmbSpatialBench | 0.843 points | 03 | 8 | 71.4% | C | |
| Multi-IF | 79.1% | 03 | 20 | 89.5% | C | |
| RefCOCO-avg | 0.924 points | 03 | 7 | 66.7% | C | |
| SUNRGBD | 0.349 points | 03 | 4 | 33.3% | C | |
| VisuLogic | 0.344 points | 03 | 3 | 0.0% | C | |
| WritingBench | 86.7% | 03 | 15 | 85.7% | C | |
| DocVQAtest | 96.5% | 04 | 11 | 70.0% | C | |
| Hypersim | 0.11 points | 04 | 4 | 0.0% | C | |
| OCRBench-V2 (en) | 66.8% | 04 | 12 | 72.7% | C | |
| ScreenSpot | 95.4% | 04 | 16 | 80.0% | C | |
| CC-OCR | 81.5% | 05 | 18 | 76.5% | C | |
| Creative Writing v3 | 85.7% | 05 | 13 | 66.7% | C | |
| MMLongBench-Doc | 0.562 points | 05 | 5 | 0.0% | C | |
| MMLU | 90.6% | 05 | 100 | 96.0% | C | |
| ZEROBench-Sub | 0.277 points | 05 | 5 | 0.0% | C | |
| CountBench | 0.937 points | 06 | 6 | 0.0% | C | |
| Hallusion Bench | 66.7% | 06 | 16 | 66.7% | C | |
| MM-MT-Bench | 8.5 points | 06 | 17 | 68.8% | C | |
| VideoMME w/o sub. | 79.0% | 06 | 10 | 44.4% | C | |
| BFCL-v3 | 71.9% | 07 | 19 | 66.7% | C | |
| MMBench-V1.1 | 90.6% | 07 | 18 | 64.7% | C | |
| MMStar | 78.7% | 07 | 22 | 71.4% | C | |
| MathVista-Mini | 85.8% | 08 | 23 | 68.2% | C | |
| MMLU-Redux | 93.7% | 08 | 48 | 85.1% | C | |
| BLINK | 67.1% | 09 | 13 | 33.3% | C | |
| MLVU | 83.8% | 09 | 10 | 11.1% | C | |
| SimpleVQA | 0.613 points | 09 | 13 | 33.3% | C | |
| ZEROBench | 0.04 points | 09 | 9 | 0.0% | C | |
| Include | 80.0% | 10 | 31 | 70.0% | C | |
| MMLU-ProX | 80.6% | 10 | 32 | 71.0% | C | |
| ODinW | 43.2% | 10 | 16 | 40.0% | C | |
| OSWorld | 38.1% | 11 | 20 | 47.4% | C | |
| RealWorldQA | 81.3% | 11 | 26 | 60.0% | C | |
| LVBench | 63.6% | 12 | 24 | 52.2% | C | |
| ERQA | 52.5% | 13 | 23 | 45.5% | C | |
| HMMT25 | 77.4% | 13 | 25 | 50.0% | C | |
| OCRBench | 87.5% | 13 | 22 | 42.9% | C | |
| SuperGPQA | 64.3% | 13 | 34 | 63.6% | C | |
| AI2D | 89.2% | 14 | 32 | 58.1% | C | |
| MathVision | 74.6% | 14 | 32 | 58.1% | C | |
| ScreenSpot Pro | 61.8% | 14 | 25 | 45.8% | C | |
| VideoMMMU | 80.0% | 17 | 26 | 36.0% | C | |
| SimpleQA | 44.4% | 18 | 46 | 62.2% | C | |
| IFEval | 88.2% | 27 | 65 | 59.4% | C | |
| MMLU-Pro | 83.8% | 28 | 129 | 78.9% | C | |
| LiveCodeBench v6 | 70.1% | 29 | 53 | 46.1% | C | |
| CharXiv-R | 66.1% | 33 | 48 | 31.9% | C | |
| MMMU-Pro | 69.3% | 37 | 66 | 44.6% | C | |
| AIME 2025 | 89.7% | 49 | 114 | 57.5% | C | |
| Humanity's Last Exam | 13.6% | 76 | 93 | 18.5% | C |
Qwen3 VL 235B A22B Thinking Arena Results
Preference and agent-evaluation results for the default version.
| vision | overall | 70 | 1208.2 | 2,347 | N/A | |
| vision style control | overall | 77 | 1189.8 | 2,347 | N/A | |
| text | overall | 130 | 1400.8 | 7,987 | N/A | |
| text style control | overall | 144 | 1395.3 | 7,987 | N/A |
Qwen3 VL 235B A22B Thinking Runtime Performance
Provider-specific output speed and catalog latency for Qwen3 VL 235B A22B Thinking. Runtime does not affect the capability score.
No runtime data
Browse runtime rankingsNo provider-specific speed or latency record is linked to the default version yet.
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Qwen3 VL 235B A22B Thinking Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- N/A
- Official provider
- N/A
- Lowest third-party
- From $0.40 input, $4 output per 1M via OpenRouter
- Tracked offerings
- 4
| OpenRouter | qwen/qwen3-vl-235b-a22b-thinking | global | $0.40 | $4 | 131.1K | |
| NanoGPT | qwen3-vl-235b-a22b-thinking | global | $0.50 | $6 | 32.8K | |
| LLM Gateway | qwen3-vl-235b-a22b-thinking | global | $0.98 | $3.95 | 131.1K | |
| NovitaAI | qwen/qwen3-vl-235b-a22b-thinking | global | $0.98 | $3.95 | 131.1K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Qwen3 VL 235B A22B Thinking Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
| Qwen3 VL 235B A22B Thinking | 45.8 | 236B | 262.1K | 262.1K | No | Apache 2.0 |
What is Qwen3 VL 235B A22B Thinking?
Key information about Qwen3 VL 235B A22B Thinking and its available data.
Qwen3-VL-235B-A22B-Thinking is the most powerful vision-language model in the Qwen series, featuring 236B parameters with MoE architecture for reasoning-enhanced multimodal understanding.
io/HTML/CSS/JS from images/videos), Advanced Spatial Perception (2D grounding and 3D grounding for spatial reasoning and embodied AI), Long Context & Video Understanding (native 256K context expandable to 1M, handles hours-long video with second-level indexing), Enhanced Multimodal Reasoning (excels in STEM/Math with causal analysis), Upgraded Visual Recognition (celebrities, anime, products, landmarks, flora/fauna), and Expanded OCR (32 languages, robust in low light/blur/tilt).
Architecture innovations include Interleaved-MRoPE for positional embeddings, DeepStack for multi-level ViT feature fusion, and Text-Timestamp Alignment for precise video temporal modeling.
Data as of 2026-08-11.
Qwen3 VL 235B A22B Thinking vs nearby models
Open a comparison with the three ranked models immediately above and below this model.
Models similar to Qwen3 VL 235B A22B Thinking
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
AC
Qwen3 VL 235B A22B
Alibaba Cloud / Qwen Team
AC
Qwen3 235B A22B Thinking
Alibaba Cloud / Qwen Team
AC
Qwen3 235B A22B
Alibaba Cloud / Qwen Team
AC
Qwen3.5 9B
Alibaba Cloud / Qwen Team
AC
Qwen3 Next 80B A3B Thinking
Alibaba Cloud / Qwen Team
AC
Qwen3 Max
Alibaba Cloud / Qwen Team
FAQ
Common questions about Qwen3 VL 235B A22B Thinking.
When was Qwen3 VL 235B A22B Thinking released?
Qwen3 VL 235B A22B Thinking's default version was released on Sep 22, 2025.
How much does Qwen3 VL 235B A22B Thinking cost?
No official standard PAYG price is currently available for Qwen3 VL 235B A22B Thinking. The lowest tracked third-party offer starts at $0.40 input and $4 output via OpenRouter.
Who created Qwen3 VL 235B A22B Thinking?
Qwen3 VL 235B A22B Thinking was created by Alibaba Cloud / Qwen Team.
What is the context window for Qwen3 VL 235B A22B Thinking?
The default version has a 262.1K token context window.
Is Qwen3 VL 235B A22B Thinking open weight?
No. The default version is not marked as having publicly available weights.
How many API providers offer Qwen3 VL 235B A22B Thinking?
4 provider offerings are linked to the default version.
What models should I compare Qwen3 VL 235B A22B Thinking with?
Nearby ranked alternatives include Grok 4.1, GPT-OSS-20B, Claude Opus 4.