Mistral OCR 3 is designed to extract text and embedded images from a wide range of documents with exceptional fidelity. It supports markdown output enriched with HTML-based table reconstruction, enabling downstream systems to understand not just document content, but also structure.

Key features include accurate interpretation of cursive, mixed-content annotations, and handwritten text layered over printed forms. The model excels at detecting boxes, labels, handwritten entries, and dense layouts on invoices, receipts, compliance forms, and government documents. It reconstructs table structures with headers, merged cells, multi-row blocks, and column hierarchies, outputting HTML table tags with colspan/rowspan to fully preserve layout.

The model achieves breakthrough performance with a 74% overall win rate over Mistral OCR 2 on forms, scanned documents, complex tables, and handwriting. It is significantly more robust to compression artifacts, skew, distortion, low DPI, and background noise compared to previous generations.

Benefits include automated parsing of forms, invoices, and operational documents, digitization of handwritten or historical documents, and extraction of clean text from technical and scientific reports. Early customers use it to process invoices into structured fields, digitize company archives, and improve enterprise search.

The product targets developers who can integrate the model via API and users leveraging Document AI UI for parsing documents into text or structured JSON. It's ideal for both high-volume enterprise pipelines and interactive document workflows.