Mistral OCR
Mistral OCR
Developer API that extracts text, tables and structure from PDFs and images into Markdown
Overview
Mistral OCR is a document-reading service from Mistral AI, offered to developers through an API. The documentation describes a Document AI OCR processor that extracts text and structured content from PDF documents and images, preserving document structure and hierarchy. Tables can be returned inline or as separate Markdown or HTML tables, headers and footers can be extracted on request, and the response includes page-by-page Markdown, image bounding boxes and metadata. Related features in the same docs cover document annotations and asking questions about a document. The commercial terms describe Mistral AI as a French limited joint-stock corporation.
This is a developer tool rather than a consumer app. There is no upload-and-click interface described on the pages read, and use means calling an API with a key. The pricing FAQ says OCR is priced per 1,000 pages, unlike most Mistral models, which are priced per million tokens, and it mentions a free plan and a Pro plan for the company’s other products. The actual per-page figure was not readable on the pricing page on 16 November 2026, so check the current API price before planning volume.
Safety note: documents sent to an API are processed by a third party, so read the data-processing addendum and privacy policy before sending contracts, identity documents or health records. As with any OCR, recognition errors in numbers and tables are possible, so verify extracted figures against the original.
At a glance
Key Features
• Preserves document structure and hierarchy
• Tables as Markdown or HTML
• Header and footer extraction
• Page-by-page Markdown with image bounding boxes
• Document annotations and document question answering
Visual Walkthrough
Best For
• Data teams digitising PDFs
• Companies feeding documents to AI systems
• Researchers converting scanned papers
• People who need translation as well as OCR
• Anyone needing a published flat per-page price
• Users who cannot send documents to an API
Use cases
2. Extract tables as HTML or Markdown.
3. Pull headers and footers separately.
4. Ask questions about a document.
5. Prepare documents for search or AI assistants.
Pricing
OCR is priced per 1,000 pages. The figure was not readable on the pricing page.
Other Mistral models are priced per million tokens.
A free and a Pro plan exist for the vendor's chat products.
Prices change often. Check the official pricing page for current prices.
Pros and cons
• Header and footer handling
• Documented API
• Named company, Mistral AI
• Needs developer skills
• Documents are processed by a third party
• OCR errors possible in numbers and tables
• No translation included
Integrations
• Document question answering
• Part of the wider Mistral platform
FAQ
A: The pricing FAQ says OCR is priced per 1,000 pages.
Q: Is it an app?
A: It is a developer API, documented as a Document AI OCR processor.
Q: Who operates it?
A: The commercial terms name Mistral AI, a French limited joint-stock corporation.
Alternatives
DeepL: translation of text and documents; see its listing on this site.
ABBYY FineReader PDF: OCR and PDF editing on the desktop.
Google Cloud Document AI: OCR as a cloud service.












