Mistral OCR 4

Mistral AI has released OCR 4, a document parsing system that extracts text with bounding boxes, block classification, and confidence scores while supporting 170 languages. The model outperforms competing OCR and document-AI systems in independent evaluations, can be self-hosted in a single container, and integrates with Mistral's Search Toolkit for enterprise search and RAG applications. It provides structured outputs designed for semantic chunking, agent workflows, and data pipelines while maintaining data privacy through on-premises deployment options.