Mistral AI offers Mistral OCR 4, a document understanding model that extracts and structures text, tables, and layout information from documents in 170 languages, available via API, Mistral Studio, and cloud platforms including Microsoft Azure and Google Cloud.

Mistral OCR 4, released in mid-2026, provides bounding-box extraction, block classification, and inline confidence scores alongside text. It supports common enterprise document formats (PDF, DOC, PPT) and is designed for use in retrieval-augmented generation (RAG) pipelines, enterprise search, and agentic workflows. Self-hosted deployment in a single container is available for enterprise customers requiring data residency or regulatory compliance. Earlier versions (OCR 1 and OCR 3) preceded OCR 4, which Mistral states supersedes them.