Mistral OCR
SOTA document-extraction API (OCR 4) returning bounding boxes, block classification and confidence scores across 170 languages; the same endpoint's Document AI mode adds a no-code, schema-driven layer on top for structured extraction; also runs fully self-hosted for data sovereignty.
Find alternatives to Mistral OCR
What it does
- Document extraction
Features 5
- Bounding boxesReturns bounding boxes that localize text for in-context highlighting and reliable data pipelines.
- Block classificationClassifies blocks by type such as titles, tables, equations and signatures.
- Inline confidence scoresGenerates confidence scores per page and per word.
- Single-container self-hostingRuns in a single container for fully self-hosted deployments, offered to enterprise customers.
- Search Toolkit integrationServes as an ingestion component of Mistral's open-source Search Toolkit for RAG and enterprise search.
Sources
- Pricing
-
OCR: $4 per 1,000 pages ($2 with the Batch API); Document AI (same endpoint, additional parameters): $5 per 1,000 pages.
- Description
- What it does
- Sold within
-
Own announcement page states both prices explicitly: 'Mistral OCR 4 through the API is priced at $4 per 1,000 pages... Document AI is priced at $5 per 1,000 pages.'
- Name
- URL
- Interfaces
-
Official documentation · 12 Oct 2026
Page: 'Developers integrate the model via API'.
- Modalities
-
Official documentation · 1 Oct 2026
Page: 'extracts and structures content from a wide range of documents / Alongside the extracted text, OCR 4 returns bounding boxes, typed-block classification'.
Also from Mistral AI
-
Mistral Vibe Application
Assistant for chat, web search, document analysis and image generation, built on Mistral's own models.
-
Mistral Studio Model API
Platform for building, deploying and governing agentic AI systems and applications on Mistral models, with workflows, connectors, agents, OCR and audio, billed per token on the API.
-
Mistral Compute Platform
Frontier-grade infrastructure and orchestration for training and serving models at scale.
-
Mistral Forge Platform
Turn institutional knowledge into custom enterprise LLMs without managing the infrastructure; also generates training samples for your workflows and measures model quality against enterprise outcomes.
-
Mistral Speech Application
Enterprise voice-AI solution set: real-time voice agents, text-to-speech/voice cloning (Voxtral TTS) and speech-to-text/diarization (Voxtral Realtime, Voxtral Mini Transcribe 2), open-weight and self-hostable.
-
Mistral Vibe Code AI agent
Agentic coding across terminal, IDE, web and background — multi-file orchestration, codebase-aware completion, async agents and native IDE extensions, built on Mistral Medium/Devstral/Codestral; also writes unit and integration tests matching existing patterns.
Open Mistral AI in the directory
Something wrong here? Send a correction — quote this product id: mistral-ocr.