Mistral OCR 4.1 reaches General Availability

Mistral has moved OCR 4.1, its document-intelligence model, out of preview and into General Availability, according to the company's changelog.

What's new

Mistral's changelog states plainly: "OCR 4.1 (mistral-ocr-4-1) is now Generally Available." The model card for OCR 4.1 describes it as "our latest OCR service powering our Document AI stack, with native paragraph-level bounding box extraction, structural block labels, and block-level confidence scores."

The model sits in Mistral's Premier tier and supports paragraph-level bounding boxes, structural block labels, block-level confidence scores, structured annotations, and batch processing. Pricing is $4 per 1,000 pages for plain OCR and $5 per 1,000 pages for annotated output, with Mistral's standard batch discount available on the OCR path.

Context

OCR 4.1 builds on OCR 4, which Mistral introduced as a state-of-the-art document intelligence model earlier this year with paragraph-level bounding boxes and structural block labels. The 4.1 update added block-level confidence scores and batching on top of that foundation, and had been available as a public preview since mid-July before today's move to General Availability. The model is offered both through Mistral's API on a per-page basis and as a self-hosted single container for enterprise customers.

Why it matters

Document intelligence — pulling structured, verifiable text and layout out of scanned or messy PDFs — has become a competitive front among model providers serving enterprise customers, alongside efforts like Cohere's recently released Parse model for the same use case. Moving OCR 4.1 to General Availability signals Mistral is ready to put the model behind production SLAs rather than preview caveats, and the added confidence scores give enterprises a way to programmatically flag low-confidence extractions for human review — a common requirement in regulated document-processing workflows like finance, healthcare, and legal.