PyyanThe AI Index

Pyyan / OCR & Document AI / Qwen3-VL

Qwen3-VL

A general vision model that happens to lead OCR benchmarks.

currentAlibaba ↗Verified 2026-08-13

32 langs

recognised

Specifications

OmniDocBench~93
Open weightsYes
HandlesDocuments, charts, video
LicenceApache 2.0
KindVision language model

Official links

Alibaba official site

What matters

  • Not an OCR model, and beats most of them
  • Sizes from 2B to 235B

Best for one model for vision and documents.

Compare with

All 5 side by side →

Qwen3-VL vs GLM-OCR →Qwen3-VL vs dots.ocr →Qwen3-VL vs DeepSeek-OCR →Qwen3-VL vs Mistral OCR →Qwen3-VL vs olmOCR →

Also in OCR & Document AI

GLM-OCR · Zhipu AIdots.ocr · XiaohongshuDeepSeek-OCR · DeepSeekMistral OCR · Mistral AIolmOCR · Allen Institute for AI