AC
Qwen3 VL 4B Thinking
Alibaba Cloud / Qwen Team
Alibaba Cloud / Qwen Team model product
Qwen2.5 VL 72B
5-VL is the new flagship vision-language model of Qwen, significantly improved from Qwen2-VL.
Updated Aug 12, 2026. Default version: Qwen2.5 VL 72B Instruct
LLMBoard score22.7Qwen2.5 VL 72B Instruct
Coverage0%26 benchmark families
Context window131.1KTokens
Official input priceN/AOfficial price unavailable
Qwen2.5 VL 72B Specifications
Technical details for the model's default version.
- Version
- Qwen2.5 VL 72B Instruct
- Released
- Jan 26, 2025
- Knowledge cutoff
- Apr 1, 2024
- Parameters
- 72B
- Context window
- 131.1K
- Max output
- 8.2K
- Inputs
- image, text
- Outputs
- text
- Open weights
- Yes
- License
- tongyi-qianwen
Qwen2.5 VL 72B Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
No capability profile
No version of this model currently has enough benchmark coverage for a score.
Qwen2.5 VL 72B Benchmark Results
Benchmark scores for Qwen2.5 VL 72B Instruct.
| AITZ_EM | 83.2% | 01 | 3 | 100.0% | C | |
| Android Control Low_EM | 93.7% | 01 | 3 | 100.0% | C | |
| DocVQA | 96.4% | 01 | 26 | 100.0% | C | |
| MMBench-Video | 2.0% | 01 | 3 | 100.0% | C | |
| MMVet | 76.2% | 01 | 2 | 100.0% | C | |
| PerceptionTest | 73.2% | 01 | 2 | 100.0% | C | |
| TempCompass | 74.8% | 01 | 2 | 100.0% | C | |
| Android Control High_EM | 67.4% | 02 | 3 | 50.0% | C | |
| EgoSchema | 76.2% | 02 | 9 | 87.5% | C | |
| MMBench | 88.0% | 02 | 9 | 87.5% | C | |
| MobileMiniWob++_SR | 68.0% | 02 | 2 | 0.0% | C | |
| ChartQA | 89.5% | 03 | 24 | 91.3% | C | |
| AndroidWorld_SR | 35.0% | 06 | 8 | 28.6% | C | |
| MLVU-M | 74.6% | 08 | 8 | 0.0% | C | |
| VideoMME w/o sub. | 73.3% | 08 | 10 | 22.2% | C | |
| OCRBench | 88.5% | 10 | 22 | 57.1% | C | |
| MVBench | 70.4% | 11 | 17 | 37.5% | C | |
| OCRBench-V2 (en) | 61.5% | 11 | 12 | 9.1% | C | |
| CC-OCR | 79.8% | 12 | 18 | 35.3% | C | |
| ScreenSpot | 87.1% | 13 | 16 | 20.0% | C | |
| Hallusion Bench | 55.2% | 15 | 16 | 6.7% | C | |
| MMStar | 70.8% | 15 | 22 | 33.3% | C | |
| AI2D | 88.4% | 16 | 32 | 51.6% | C | |
| MathVista-Mini | 74.8% | 16 | 23 | 31.8% | C | |
| OSWorld | 8.8% | 19 | 20 | 5.3% | C | |
| LVBench | 47.3% | 21 | 24 | 13.0% | C | |
| ScreenSpot Pro | 43.6% | 23 | 25 | 8.3% | C | |
| MathVision | 38.1% | 29 | 32 | 9.7% | C | |
| MMMU | 70.2% | 31 | 63 | 51.6% | C | |
| MMMU-Pro | 51.1% | 58 | 66 | 12.3% | C |
Qwen2.5 VL 72B Arena Results
Preference and agent-evaluation results for the default version.
| vision | overall | 108 | 1107.6 | 3,768 | N/A | |
| vision style control | overall | 109 | 1121.8 | 3,768 | N/A |
Qwen2.5 VL 72B Runtime Performance
Provider-specific output speed and catalog latency for Qwen2.5 VL 72B Instruct. 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.
Qwen2.5 VL 72B Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- N/A
- Official provider
- N/A
- Lowest third-party
- N/A
- Tracked offerings
- 0
No provider prices
The default version has no 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.
Qwen2.5 VL 72B Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
| Qwen2.5 VL 72B Instruct | 22.7 | 72B | 131.1K | 8.2K | Yes | tongyi-qianwen |
What is Qwen2.5 VL 72B?
Key information about Qwen2.5 VL 72B and its available data.
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.
Data as of 2026-08-11.
Qwen2.5 VL 72B vs nearby models
Open a comparison with the three ranked models immediately above and below this model.
Models similar to Qwen2.5 VL 72B
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
AC
Qwen2.5 72B
Alibaba Cloud / Qwen Team
AC
Qwen3 VL 8B
Alibaba Cloud / Qwen Team
AC
QwQ 32B
Alibaba Cloud / Qwen Team
AC
Qwen2 VL 72B
Alibaba Cloud / Qwen Team
AC
Qwen3 VL 4B
Alibaba Cloud / Qwen Team
FAQ
Common questions about Qwen2.5 VL 72B.
When was Qwen2.5 VL 72B released?
Qwen2.5 VL 72B's default version was released on Jan 26, 2025.
How much does Qwen2.5 VL 72B cost?
No official standard PAYG price is currently available for Qwen2.5 VL 72B.
Who created Qwen2.5 VL 72B?
Qwen2.5 VL 72B was created by Alibaba Cloud / Qwen Team.
What is the context window for Qwen2.5 VL 72B?
The default version has a 131.1K token context window.
Is Qwen2.5 VL 72B open weight?
Yes. The default version is marked as open weight under tongyi-qianwen.
How many API providers offer Qwen2.5 VL 72B?
No provider offering is currently linked to the default version.
What models should I compare Qwen2.5 VL 72B with?
Nearby ranked alternatives include GPT-4o, Qwen3 VL 4B Thinking, Llama 3.3 Nemotron Super 49B.