← All models

Qwen-VL-OCR

Published Publisher Alibabaqwen-vl-ocr-2025-04-13

Model Positioning

A visual understanding and OCR model trained on Qwen-VL, using a unified model to handle image and text recognition, parsing, and processing tasks. This version is a snapshot from April 13, 2025. (docs.bailian.console.aliyun.com)

Capabilities and Uses

It includes six types of tasks: general recognition, document parsing, table parsing, information extraction, multilingual recognition, and formula recognition. It can be used to extract and organize text information from documents, tables, and other images. (docs.bailian.console.aliyun.com)

Limitations and Service Status

  • Hallucinations may occur when text in an image is too small or the resolution is too low; answer accuracy is not guaranteed for questions outside text extraction. This generation and earlier versions process only the most recent message and do not retain multi-turn context. (docs.modelstudio.console.alibabacloud.com)
  • An Alibaba Cloud international site notice lists this model among those scheduled for retirement on October 10, 2026, and states that inference will no longer be supported after retirement; the actual change date is subject to the platform. Specific availability should be verified with the relevant platform. (alibabacloud.com)

Release Date

April 23, 2025; the model snapshot date is April 13, 2025. (docs.bailian.console.aliyun.com)

Pricing

The historical Singapore region prices listed in the Alibaba Cloud international site retirement notice are: input 0.717 USD/million tokens, output 0.717 USD/million tokens. These prices come from the retirement notice and should not be considered current prices for available calls. Query date: November 18, 2026. (alibabacloud.com)

Official Sources