[AI Summary]: DeepSeek-OCR is an innovative optical compression technology that addresses the high computational costs of processing long contexts in Large Language Models (LLMs) by converting documents to high-resolution images and compressing them into significantly fewer vision tokens (10x compression while maintaining 97% accuracy). Unlike traditional OCR, it understands context and structure including tables, charts, formulas and layouts, supports about 100 languages, and can output structured formats like Markdown or HTML.
- Developer: DeepSeek AI
- License: MIT License
- Platform: GitHub
- Languages: Supports ~100 languages