Introducing Mistral OCR 3

Mistral OCR 3 is a new model designed for high-fidelity text and image extraction from a wide range of documents. It demonstrates a 74% overall win rate over the previous version, showing breakthrough performance on forms, complex tables, and handwriting. The model supports markdown output with HTML-based table reconstruction to preserve document structure. Developers can integrate it via an API, and users can access it through the Document AI Playground for parsing documents into text or structured JSON.