Alibaba Cloud / Qwen Team model product
Qwen2.5-VL is a vision-language model from the Qwen family. Key enhancements include visual understanding (objects, text, charts, layouts), visual agent capabilities (tool use, computer/phone control), long video comprehension with event pinpointing, visual localization (bounding boxes/points), and structured output generation.
Updated Aug 10, 2026. Default version: Qwen2.5 VL 32B 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 Qwen2.5 VL 32B Instruct.
| Android Control High_EM | 0.7 | 1 | 3 | 100.0% | C | |
| InfoVQA | 0.8 | 1 | 9 | 100.0% | C | |
| AITZ_EM | 0.8 | 2 | 3 | 50.0% | C | |
| Android Control Low_EM | 0.9 | 2 | 3 | 50.0% | C | |
| MMBench-Video | 0.0 | 2 | 3 | 50.0% | C | |
| DocVQA | 0.9 | 6 | 26 | 80.0% | C | |
| AndroidWorld_SR | 0.2 | 8 | 8 | 0.0% | C | |
| MBPP | 0.8 | 8 | 37 | 80.6% | C | |
| OCRBench-V2 (zh) | 0.6 | 8 | 11 | 30.0% | C | |
| VideoMME w sub. | 0.8 | 8 | 10 | 22.2% | C | |
| VideoMME w/o sub. | 0.7 | 9 | 10 | 11.1% | C | |
| CharadesSTA | 0.5 | 11 | 12 | 9.1% | C | |
| ScreenSpot | 0.9 | 11 | 16 | 33.3% | C | |
| OCRBench-V2 (en) | 0.6 | 12 | 12 | 0.0% | C | |
| HumanEval | 0.9 | 13 | 76 | 84.0% | C | |
| CC-OCR | 0.8 | 15 | 18 | 17.6% | C | |
| MathVista-Mini | 0.7 | 17 | 23 | 27.3% | C | |
| MMStar | 0.7 | 17 | 22 | 23.8% | C | |
| MATH | 0.8 | 18 | 71 | 75.7% | C | |
| LVBench | 0.5 | 20 | 24 | 17.4% | C | |
| OSWorld | 0.1 | 20 | 20 | 0.0% | C | |
| ScreenSpot Pro | 0.4 | 23 | 24 | 4.3% | C | |
| MathVision | 0.4 | 28 | 32 | 12.9% | C | |
| MMMU | 0.7 | 33 | 63 | 48.4% | C | |
| MMMU-Pro | 0.5 | 58 | 65 | 10.9% | C | |
| MMLU | 0.8 | 70 | 100 | 30.3% | C | |
| MMLU-Pro | 0.7 | 88 | 129 | 32.0% | C | |
| GPQA | 0.5 | 188 | 233 | 19.4% | C |
Preference and agent-evaluation signals from published Arena datasets.
| vision | overall | 97 | 1152.5 | 1,490 | N/A | |
| vision | english | 98 | 1159.3 | 870 | N/A | |
| vision style control | overall | 111 | 1119.4 | 1,490 | N/A | |
| vision style control | english | 111 | 1124.5 | 870 | 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.
| IO.NET | Qwen/Qwen2.5-VL-32B-Instruct | global | $0.05 | $0.22 | 32K | |
| Meganova | Qwen/Qwen2.5-VL-32B-Instruct | global | $0.20 | $0.60 | 16.4K | |
| LLM Gateway | qwen2-5-vl-32b-instruct | global | $1.4 | $4.2 | 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.
| Qwen2.5 VL 32B Instruct | 51.5 | 33.5B | N/A | N/A | No | Apache 2.0 |
A concise description based on the published model registry.
Qwen2.5-VL is a vision-language model from the Qwen family. Key enhancements include visual understanding (objects, text, charts, layouts), visual agent capabilities (tool use, computer/phone control), long video comprehension with event pinpointing, visual localization (bounding boxes/points), and structured output generation.
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 Qwen2.5 VL 32B.
Qwen2.5 VL 32B's default version was released on Feb 28, 2025.
No official standard PAYG price is currently available for Qwen2.5 VL 32B. The lowest tracked third-party offer starts at $0.05 input and $0.22 output via IO.NET.
Qwen2.5 VL 32B is published under Alibaba Cloud / Qwen Team in the model registry.
The current registry does not publish a context window for the default version.
No. The default version is not marked as having publicly available weights.
3 published provider offerings are linked to the default version.
Nearby ranked alternatives include Gemini 2.0 Flash Lite, Llama 3.1 70B, Qwen3 VL 32B Thinking.