Alibaba Cloud / Qwen Team model product
An instruction-tuned, large multimodal model that excels at visual understanding and step-by-step reasoning. It supports image and video input, with dynamic resolution handling and improved positional embeddings (M-ROPE), enabling advanced capabilities such as complex problem solving, multilingual text recognition in images, and agent-like interactions in video contexts.
Updated Aug 10, 2026. Default version: Qwen2-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-VL-72B-Instruct.
| EgoSchema | 0.8 | 1 | 9 | 100.0% | C | |
| MMVetGPT4Turbo | 0.7 | 1 | 1 | 100.0% | C | |
| MTVQA | 0.3 | 1 | 1 | 100.0% | C | |
| TextVQA | 0.9 | 1 | 15 | 100.0% | C | |
| VCR_en_easy | 0.9 | 1 | 1 | 100.0% | C | |
| DocVQAtest | 1.0 | 3 | 11 | 80.0% | C | |
| MMBench | 0.9 | 4 | 9 | 62.5% | C | |
| MMMUval | 0.6 | 4 | 4 | 0.0% | C | |
| ChartQA | 0.9 | 6 | 24 | 78.3% | C | |
| MVBench | 0.7 | 6 | 17 | 68.8% | C | |
| InfoVQAtest | 0.8 | 8 | 12 | 36.4% | C | |
| OCRBench | 0.9 | 12 | 22 | 47.6% | C | |
| RealWorldQA | 0.8 | 15 | 26 | 44.0% | C | |
| MathVista-Mini | 0.7 | 19 | 23 | 18.2% | C | |
| MMMU-Pro | 0.5 | 59 | 65 | 9.4% | C |
Preference and agent-evaluation signals from published Arena datasets.
| vision style control | chinese | 85 | 1107.8 | 291 | N/A | |
| vision | chinese | 87 | 1066.5 | 291 | N/A | |
| vision style control | overall | 118 | 1085.5 | 5,937 | N/A | |
| vision | overall | 119 | 1047.1 | 5,937 | N/A | |
| vision style control | english | 120 | 1085.6 | 3,230 | N/A | |
| vision | english | 121 | 1054.9 | 3,230 | 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.
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-VL-72B-Instruct | 25.6 | 73.4B | N/A | N/A | No | tongyi-qianwen |
A concise description based on the published model registry.
An instruction-tuned, large multimodal model that excels at visual understanding and step-by-step reasoning. It supports image and video input, with dynamic resolution handling and improved positional embeddings (M-ROPE), enabling advanced capabilities such as complex problem solving, multilingual text recognition in images, and agent-like interactions in video contexts.
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 VL 72B.
Qwen2 VL 72B's default version was released on Aug 29, 2024.
No official standard PAYG price is currently available for Qwen2 VL 72B.
Qwen2 VL 72B 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.
No published provider offering is currently linked to the default version.
Nearby ranked alternatives include Llama 4 Maverick, DeepSeek-V3, QwQ 32B.