Mistral OCR 4 : SOTA OCR for Document Intelligence

Mistral released OCR 4, a document-parsing model supporting 170 languages with bounding boxes, block classification, and confidence scores, priced at $4 per 1,000 pages.

• OCR 4 scores 85.20 on OlmOCRBench, 93.07 on OmniDocBench, and wins human preference comparisons with an average 72% win rate.
• It is available via API through Mistral Studio, Amazon SageMaker, Microsoft Foundry, and soon Snowflake Parse Document, with self-hosting in a single container.
• The Batch API discount lowers the price to $2 per 1,000 pages, and Document AI is priced at $5 per 1,000 pages.
• It is integrated with Mistral Search Toolkit, now in public preview, as an ingestion component for RAG and enterprise search.

Teams building RAG, enterprise search, and agentic document pipelines can now get structured, localized, confidence-scored OCR output that can be self-hosted.

Sources

• Mistral: Mistral OCR 4 : SOTA OCR for Document Intelligence

Read this as text

Back to the AI news