Alibaba Cloud / Qwen Team model product
Qwen3-VL is a large multimodal model that unifies vision, language, and reasoning to achieve human-level perception and cognition across text, images, and video. Built on a 235B-parameter architecture, it integrates early joint training of visual and textual modalities for strong language grounding. The model supports up to a 1 million-token context window and excels at visual understanding, spatial reasoning, long video comprehension, and tool-based interaction. It can generate code from images, perform precise 2D/3D object grounding, and operate digital interfaces like a visual agent. The “Instruct” version rivals Gemini 2.5 Pro in perception benchmarks, while the “Thinking” version leads in multimodal reasoning and STEM tasks. With multilingual OCR, creative writing, and fine-grained scene interpretation, Qwen3-VL establishes a new open-source frontier for integrated vision-language intelligence.
Updated Aug 10, 2026. Default version: Qwen3 VL 235B A22B Instruct
Structured fields from the published default version.
This profile uses the latest version under this unique model that has a calculated LLMBoard score. Arena and price are excluded.
Published benchmark records for the scored version Qwen3 VL 235B A22B Instruct.
| CharadesSTA | 0.6 | 1 | 12 | 100.0% | C | |
| DocVQAtest | 1.0 | 1 | 11 | 100.0% | C | |
| CC-OCR | 0.8 | 2 | 18 | 94.1% | C | |
| Creative Writing v3 | 0.9 | 2 | 13 | 91.7% | C | |
| MMMUval | 0.8 | 2 | 4 | 66.7% | C | |
| BLINK | 0.7 | 3 | 13 | 83.3% | C | |
| CSimpleQA | 0.8 | 3 | 7 | 66.7% | C | |
| InfoVQAtest | 0.9 | 3 | 12 | 81.8% | C | |
| LiveCodeBench v5 | 0.6 | 3 | 9 | 75.0% | C | |
| MultiPL-E | 0.9 | 3 | 13 | 83.3% | C | |
| OCRBench | 0.9 | 3 | 22 | 90.5% | C | |
| OCRBench-V2 (en) | 0.7 | 3 | 12 | 81.8% | C | |
| OCRBench-V2 (zh) | 0.6 | 3 | 11 | 80.0% | C | |
| ScreenSpot | 1.0 | 3 | 16 | 86.7% | C | |
| AndroidWorld_SR | 0.6 | 4 | 8 | 57.1% | C | |
| ODinW | 0.5 | 4 | 16 | 80.0% | C | |
| Arena-Hard v2 | 0.8 | 5 | 16 | 73.3% | C | |
| MM-MT-Bench | 8.5 | 5 | 17 | 75.0% | C | |
| OSWorld | 0.7 | 5 | 20 | 79.0% | C | |
| VideoMME w/o sub. | 0.8 | 5 | 10 | 55.6% | C | |
| WritingBench | 0.9 | 5 | 15 | 71.4% | C | |
| LiveBench 20241125 | 0.7 | 6 | 14 | 61.5% | C | |
| MuirBench | 0.7 | 6 | 11 | 50.0% | C | |
| Multi-IF | 0.8 | 7 | 20 | 68.4% | C | |
| MLVU | 0.8 | 8 | 10 | 22.2% | C | |
| MMBench-V1.1 | 0.9 | 8 | 18 | 58.8% | C | |
| MMStar | 0.8 | 8 | 22 | 66.7% | C | |
| Include | 0.8 | 9 | 31 | 73.3% | C | |
| MathVista-Mini | 0.8 | 9 | 23 | 63.6% | C | |
| LVBench | 0.7 | 10 | 24 | 60.9% | C | |
| Hallusion Bench | 0.6 | 11 | 16 | 33.3% | C | |
| AI2D | 0.9 | 12 | 32 | 64.5% | C | |
| RealWorldQA | 0.8 | 12 | 26 | 56.0% | C | |
| ScreenSpot Pro | 0.6 | 12 | 24 | 52.2% | C | |
| SimpleQA | 0.5 | 13 | 46 | 73.3% | C | |
| BFCL-v3 | 0.7 | 14 | 19 | 27.8% | C | |
| MMLU-ProX | 0.8 | 14 | 32 | 58.1% | C | |
| ERQA | 0.5 | 15 | 23 | 36.4% | C | |
| MMLU | 0.9 | 16 | 100 | 84.8% | C | |
| SuperGPQA | 0.6 | 17 | 34 | 51.5% | C | |
| MathVision | 0.7 | 18 | 32 | 45.2% | C | |
| HMMT25 | 0.6 | 19 | 25 | 25.0% | C | |
| MMLU-Redux | 0.9 | 19 | 48 | 61.7% | C | |
| VideoMMMU | 0.7 | 20 | 26 | 24.0% | C | |
| IFEval | 0.9 | 29 | 65 | 56.3% | C | |
| CharXiv-R | 0.6 | 35 | 47 | 26.1% | C | |
| MMMU-Pro | 0.7 | 38 | 65 | 42.2% | C | |
| MMLU-Pro | 0.8 | 40 | 129 | 69.5% | C | |
| LiveCodeBench v6 | 0.5 | 41 | 53 | 23.1% | C | |
| AIME 2025 | 0.7 | 79 | 114 | 31.0% | C |
Preference and agent-evaluation signals from published Arena datasets.
| vision | captioning | 6 | 1277.8 | 143 | N/A | |
| vision style control | captioning | 12 | 1207.7 | 143 | N/A | |
| vision | creative writing | 14 | 1224.8 | 690 | N/A | |
| vision | entity recognition | 15 | 1255.6 | 147 | N/A | |
| vision style control | creative writing | 19 | 1193.5 | 690 | N/A | |
| vision style control | entity recognition | 28 | 1190.4 | 147 | N/A | |
| vision | homework | 36 | 1280.2 | 1,567 | N/A | |
| vision | chinese | 46 | 1283.9 | 714 | N/A | |
| text | korean | 49 | 1400.0 | 308 | N/A | |
| vision | english | 49 | 1251.5 | 5,092 | N/A | |
| vision style control | homework | 49 | 1260.3 | 1,567 | N/A | |
| vision | overall | 50 | 1246.8 | 12,102 | N/A | |
| vision | ocr | 50 | 1253.6 | 7,534 | N/A | |
| vision | creative writing vision | 52 | 1229.0 | 816 | N/A | |
| vision | diagram | 52 | 1255.8 | 2,659 | N/A | |
| text | french | 53 | 1457.8 | 235 | N/A | |
| text | industry business and management and financial operations | 56 | 1432.1 | 2,230 | N/A | |
| vision | humor | 56 | 1202.7 | 449 | N/A | |
| vision style control | chinese | 57 | 1243.3 | 714 | N/A | |
| text | industry legal and government | 59 | 1446.3 | 754 | N/A | |
| vision style control | english | 61 | 1218.4 | 5,092 | N/A | |
| vision style control | creative writing vision | 62 | 1197.6 | 816 | N/A | |
| vision style control | ocr | 62 | 1229.3 | 7,534 | N/A | |
| vision style control | overall | 63 | 1214.6 | 12,102 | N/A | |
| vision style control | diagram | 64 | 1238.9 | 2,659 | N/A | |
| vision style control | humor | 64 | 1157.7 | 449 | N/A | |
| text | expert | 68 | 1445.4 | 565 | N/A | |
| text | german | 76 | 1423.5 | 259 | N/A | |
| text | industry life and physical and social science | 76 | 1443.9 | 1,748 | N/A | |
| text | japanese | 76 | 1383.1 | 151 | N/A | |
| text | polish | 81 | 1417.7 | 518 | N/A | |
| text style control | french | 81 | 1453.7 | 235 | N/A | |
| text style control | korean | 81 | 1381.7 | 308 | N/A | |
| text style control | industry business and management and financial operations | 83 | 1439.0 | 2,230 | N/A | |
| text | instruction following | 84 | 1409.2 | 2,921 | N/A | |
| text | longer query | 84 | 1423.4 | 2,195 | N/A | |
| text | hard prompts english | 85 | 1440.4 | 2,944 | N/A | |
| text | multi turn | 86 | 1429.4 | 2,014 | N/A | |
| text | math | 89 | 1425.9 | 702 | N/A | |
| text style control | industry legal and government | 89 | 1438.5 | 754 | N/A | |
| text | industry medicine and healthcare | 90 | 1435.8 | 628 | N/A | |
| text | hard prompts | 91 | 1430.3 | 5,708 | N/A | |
| text | industry software and it services | 91 | 1443.5 | 4,091 | N/A | |
| text | overall | 92 | 1421.4 | 11,483 | N/A | |
| text | chinese | 94 | 1459.7 | 492 | N/A | |
| text | coding | 94 | 1440.1 | 2,314 | N/A | |
| text | exclude ties | 94 | 1409.4 | 8,102 | N/A | |
| text | non english | 94 | 1406.2 | 5,849 | N/A | |
| text style control | japanese | 96 | 1368.7 | 151 | N/A | |
| text style control | expert | 99 | 1445.9 | 565 | N/A | |
| text | english | 100 | 1430.5 | 5,634 | N/A | |
| text style control | german | 101 | 1410.5 | 259 | N/A | |
| text style control | hard prompts english | 101 | 1451.5 | 2,944 | N/A | |
| text style control | industry life and physical and social science | 102 | 1440.9 | 1,748 | N/A | |
| text | industry mathematical | 103 | 1420.8 | 587 | N/A | |
| text style control | multi turn | 103 | 1426.1 | 2,014 | N/A | |
| text | industry writing and literature and language | 104 | 1392.5 | 2,354 | N/A | |
| text factuality | hard prompts | 104 | 1445.6 | 1,303 | N/A | |
| text | spanish | 106 | 1408.3 | 405 | N/A | |
| text style control | industry software and it services | 106 | 1455.1 | 4,091 | N/A | |
| text style control | instruction following | 106 | 1413.0 | 2,921 | N/A | |
| text | russian | 107 | 1403.2 | 714 | N/A | |
| text style control | coding | 107 | 1465.1 | 2,314 | N/A | |
| text style control | polish | 107 | 1401.8 | 518 | N/A | |
| text style control | hard prompts | 108 | 1440.2 | 5,708 | N/A | |
| text style control | chinese | 109 | 1453.5 | 492 | N/A | |
| text style control | industry medicine and healthcare | 113 | 1436.4 | 628 | N/A | |
| text style control | longer query | 113 | 1425.0 | 2,195 | N/A | |
| text style control | english | 116 | 1427.0 | 5,634 | N/A | |
| text style control | spanish | 116 | 1399.6 | 405 | N/A | |
| text style control | exclude ties | 117 | 1402.3 | 8,102 | N/A | |
| text style control | overall | 118 | 1415.2 | 11,483 | N/A | |
| text | industry entertainment and sports and media | 119 | 1371.4 | 1,898 | N/A | |
| text style control | math | 119 | 1410.5 | 702 | N/A | |
| text factuality | exclude ties | 120 | 1402.7 | 2,168 | N/A | |
| text | creative writing | 121 | 1369.9 | 1,373 | N/A | |
| text factuality | non english | 122 | 1396.3 | 1,525 | N/A | |
| text style control | non english | 122 | 1397.3 | 5,849 | N/A | |
| text factuality | overall | 123 | 1410.1 | 3,099 | N/A | |
| text style control | russian | 125 | 1400.7 | 714 | N/A | |
| text style control | industry writing and literature and language | 126 | 1387.6 | 2,354 | N/A | |
| text style control | industry mathematical | 132 | 1410.1 | 587 | N/A | |
| text factuality | english | 134 | 1404.2 | 1,483 | N/A | |
| text style control | industry entertainment and sports and media | 143 | 1360.8 | 1,898 | N/A | |
| text style control | creative writing | 144 | 1361.0 | 1,373 | N/A |
Official vendor API PAYG pricing is summarized first. The table then lists individual provider offerings without treating their minimum as the official price.
| LLM Gateway | qwen3-vl-235b-a22b-instruct | global | $0.20 | $0.88 | 262.1K | |
| TensorX | qwen/qwen3-vl-235b-a22b-instruct | global | $0.21 | $1.9 | 131K | |
| OpenRouter | qwen/qwen3-vl-235b-a22b-instruct | global | $0.21 | $1.9 | 262.1K | |
| NanoGPT | qwen/Qwen3-VL-235B-A22B-Instruct | global | $0.30 | $1.2 | 128K | |
| Helicone | qwen3-vl-235b-a22b-instruct | global | $0.30 | $1.5 | 256K | |
| NovitaAI | qwen/qwen3-vl-235b-a22b-instruct | global | $0.30 | $1.5 | 131.1K | |
| Vercel AI Gateway | alibaba/qwen3-vl-235b-a22b-instruct | global | $0.40 | $1.6 | 131.1K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
All published versions linked to this unique model. The score columns identify the version used by the current overall ranking.
| Qwen3 VL 235B A22B Instruct | 42.7 | 236B | 262.1K | 262.1K | No | Apache 2.0 |
A concise description based on the published model registry.
Qwen3-VL is a large multimodal model that unifies vision, language, and reasoning to achieve human-level perception and cognition across text, images, and video. Built on a 235B-parameter architecture, it integrates early joint training of visual and textual modalities for strong language grounding. The model supports up to a 1 million-token context window and excels at visual understanding, spatial reasoning, long video comprehension, and tool-based interaction. It can generate code from images, perform precise 2D/3D object grounding, and operate digital interfaces like a visual agent. The “Instruct” version rivals Gemini 2.5 Pro in perception benchmarks, while the “Thinking” version leads in multimodal reasoning and STEM tasks. With multilingual OCR, creative writing, and fine-grained scene interpretation, Qwen3-VL establishes a new open-source frontier for integrated vision-language intelligence.
Use the benchmark, Arena and pricing sections above as separate evidence. A missing field means the current data snapshot does not support that claim.
Data snapshot: 2026-08-07. Editorial model content is not available in the backend.
Open a comparison with the three ranked models immediately above and below this model.
Recommendations prioritize the same model type and family, then the closest published LLMBoard score.
Common questions about Qwen3 VL 235B A22B.
Qwen3 VL 235B A22B's default version was released on Sep 22, 2025.
No official standard PAYG price is currently available for Qwen3 VL 235B A22B. The lowest tracked third-party offer starts at $0.20 input and $0.88 output via LLM Gateway.
Qwen3 VL 235B A22B is published under Alibaba Cloud / Qwen Team in the model registry.
The default version has a 262.1K token context window.
No. The default version is not marked as having publicly available weights.
7 published provider offerings are linked to the default version.
Nearby ranked alternatives include Llama 3.1 Nemotron 70B, Mistral Small 3.1 24B, GPT-5.4-nano.