On June 23, 2026, Mistral announced Mistral OCR 4, a document-intelligence model that returns bounding boxes, block classification (titles, tables, equations, signatures, and more), and inline confidence scores alongside extracted text. The model supports 170 languages across 10 language groups and can be deployed in a single container for self-hosting.
Mistral reports an OlmOCRBench score of 85.20 and an OmniDocBench score of 93.07, with human annotators preferring OCR 4 over competing systems at win rates averaging 72 percent. Pricing is 4 dollars per 1,000 pages via API (2 dollars with the Batch API discount) and 5 dollars per 1,000 pages via Document AI.
The model is positioned as an ingestion component for enterprise search, RAG systems, and domain-specific retrieval, and is available through Mistral Studio, Amazon SageMaker, Microsoft Foundry, and self-hosting for data-privacy-sensitive organizations.