Alibaba Cloud / Qwen Team model product
Qwen2.5-VL is the new flagship vision-language model of Qwen, significantly improved from Qwen2-VL. It excels at recognizing objects, analyzing text/charts/layouts in images, acting as a visual agent, understanding long videos (over 1 hour) with event pinpointing, performing visual localization (bounding boxes/points), and generating structured outputs from documents.
Updated Aug 10, 2026. Default version: Qwen2.5 VL 72B 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 72B Instruct.
| AITZ_EM | 0.8 | 1 | 3 | 100.0% | C | |
| Android Control Low_EM | 0.9 | 1 | 3 | 100.0% | C | |
| DocVQA | 1.0 | 1 | 26 | 100.0% | C | |
| MMBench-Video | 0.0 | 1 | 3 | 100.0% | C | |
| MMVet | 0.8 | 1 | 2 | 100.0% | C | |
| PerceptionTest | 0.7 | 1 | 2 | 100.0% | C | |
| TempCompass | 0.7 | 1 | 2 | 100.0% | C | |
| Android Control High_EM | 0.7 | 2 | 3 | 50.0% | C | |
| EgoSchema | 0.8 | 2 | 9 | 87.5% | C | |
| MMBench | 0.9 | 2 | 9 | 87.5% | C | |
| MobileMiniWob++_SR | 0.7 | 2 | 2 | 0.0% | C | |
| ChartQA | 0.9 | 3 | 24 | 91.3% | C | |
| AndroidWorld_SR | 0.3 | 6 | 8 | 28.6% | C | |
| MLVU-M | 0.7 | 8 | 8 | 0.0% | C | |
| VideoMME w/o sub. | 0.7 | 8 | 10 | 22.2% | C | |
| OCRBench | 0.9 | 10 | 22 | 57.1% | C | |
| MVBench | 0.7 | 11 | 17 | 37.5% | C | |
| OCRBench-V2 (en) | 0.6 | 11 | 12 | 9.1% | C | |
| CC-OCR | 0.8 | 12 | 18 | 35.3% | C | |
| ScreenSpot | 0.9 | 13 | 16 | 20.0% | C | |
| Hallusion Bench | 0.6 | 15 | 16 | 6.7% | C | |
| MMStar | 0.7 | 15 | 22 | 33.3% | C | |
| AI2D | 0.9 | 16 | 32 | 51.6% | C | |
| MathVista-Mini | 0.7 | 16 | 23 | 31.8% | C | |
| OSWorld | 0.1 | 19 | 20 | 5.3% | C | |
| LVBench | 0.5 | 21 | 24 | 13.0% | C | |
| ScreenSpot Pro | 0.4 | 22 | 24 | 8.7% | C | |
| MathVision | 0.4 | 29 | 32 | 9.7% | C | |
| MMMU | 0.7 | 31 | 63 | 51.6% | C | |
| MMMU-Pro | 0.5 | 57 | 65 | 12.5% | C |
Preference and agent-evaluation signals from published Arena datasets.
The default version has no published Arena rows, or its source alias has not been resolved.
Official vendor API PAYG pricing is summarized first. The table then lists individual provider offerings without treating their minimum as the official price.
The default version has no provider offering with current input or output token prices.
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 72B Instruct | 22.8 | 72B | 131.1K | 8.2K | Yes | tongyi-qianwen |
A concise description based on the published model registry.
Qwen2.5-VL is the new flagship vision-language model of Qwen, significantly improved from Qwen2-VL. It excels at recognizing objects, analyzing text/charts/layouts in images, acting as a visual agent, understanding long videos (over 1 hour) with event pinpointing, performing visual localization (bounding boxes/points), and generating structured outputs from documents.
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 72B.
Qwen2.5 VL 72B's default version was released on Jan 26, 2025.
No official standard PAYG price is currently available for Qwen2.5 VL 72B.
Qwen2.5 VL 72B is published under Alibaba Cloud / Qwen Team in the model registry.
The default version has a 131.1K token context window.
Yes. The default version is marked as open weight under tongyi-qianwen.
No published provider offering is currently linked to the default version.
Nearby ranked alternatives include Magistral Medium, Qwen3 VL 4B Thinking, Llama 3.3 70B.