← All models

Qwen-VL-OCR

Published Publisher Alibabaqwen-vl-ocr-2025-11-20

Positioning and release

Qwen-VL-OCR is a vision-language OCR model for recognizing, parsing and processing text in images. This model ID identifies the November 20, 2025 snapshot. Alibaba's release page lists it under November 20 in one section and under November 21 for the international deployment scope. The snapshot is based on the Qwen3-VL architecture and improves document parsing and text localization while reducing latency and hallucinations, according to Alibaba. (help.aliyun.com)

Capabilities and uses

  • Extract text and locate text lines with coordinates; parse documents, tables and formulas; recognize multilingual text; and correct image rotation.
  • Extract information from images of documents, tickets and certificates, making it useful for document digitization and information-entry workflows.
  • Accept image and text input and return text. Alibaba documents a maximum input length of 30,000 tokens, maximum output of 8,192 tokens and total context length of 38,192 tokens for this snapshot. (help.aliyun.com)

Limitations

The snapshot processes only the latest message and does not retain conversational context. It does not support a custom system message. Small text or low-resolution images can cause hallucinations, and accuracy is not guaranteed for questions unrelated to text extraction. PDFs should be converted to page images for this version. Alibaba's model information also lists function calling, structured-output mode, context caching and batch inference as unsupported for this snapshot. (help.aliyun.com)

Pricing

Prices below are official Model Studio standard inference prices, checked November 17, 2026. Input and output are billed separately, per 1 million tokens. Image input contributes to input-token usage. Prices depend on the purchasing site and deployment region; the CNY and USD listings are not interchangeable plans. (docs.bailian.console.aliyun.com)

  • Alibaba Cloud China site, CNY: Beijing, Virginia and Frankfurt: input CNY 0.3, output CNY 0.5. Singapore, international deployment: input CNY 0.514, output CNY 1.174. The Beijing listing states no free quota for this snapshot. Official China pricing. (docs.bailian.console.aliyun.com)
  • Alibaba Cloud international site, USD: Singapore, international deployment: input USD 0.07, output USD 0.16. Beijing, Frankfurt and Virginia: input USD 0.043, output USD 0.072. The Singapore listing states no free quota for this snapshot. Official international pricing. (alibabacloud.com)

Official sources