Mistral OCR 4is an advanced document processing technology that converts unstructured visual content into structured digital information. Instead of simply detecting characters on a page, it interprets how information is organized and what it means within the document.
It is commonly used in workflows involving PDF digitization, scanned document conversion, invoice processing, and enterprise data extraction. The system is designed to support modern AI pipelines where raw documents must be turned into clean, structured datasets.
| Category | Specification |
|---|---|
| Model Name | Mistral OCR 4 |
| Developer | Mistral AI |
| Model Type | AI-powered Optical Character Recognition (OCR) system |
| Core Function | Document text extraction, structure recognition, and layout-aware parsing |
| Input Formats | Images, scanned documents, PDFs, photographed pages |
| Output Formats | Structured text, JSON, Markdown |
Mistral OCR 4 delivers high-accuracy text extraction across a wide range of document conditions, including low-quality scans, photographed pages, and digitally compressed files. It reduces common recognition errors such as broken words, misplaced characters, and inconsistent spacing.
The system is capable of preserving visual hierarchy within documents. Tables, forms, multi-column layouts, headers, and footers are interpreted as structured components rather than flat text, allowing downstream systems to work with cleaner and more organized data.
Another key capability is multilingual recognition, enabling the processing of documents that contain multiple languages within the same file. This makes it suitable for global organizations operating across different regions and markets.
Handwritten content support is also included, allowing the system to extract meaningful text from notes, forms, and annotations that are not digitally typed. While handwriting remains more challenging than printed text, the model is optimized to improve recognition reliability in these scenarios.
Businesses use advanced OCR systems like Mistral OCR 4 to streamline document-heavy operations. This includes processing invoices, extracting contract clauses, and digitizing internal records. The goal is to reduce manual data entry while improving consistency and auditability across systems.
Legal teams often deal with large volumes of structured and semi-structured documents. A system like Mistral OCR 4 is particularly relevant in this space because it supports accurate extraction of clauses, references, and formatting hierarchies that are critical in legal interpretation.
Academic and research environments benefit from OCR systems capable of preserving citations, section structures, and embedded references. Mistral OCR 4 is often associated with workflows that convert scanned research materials into searchable and machine-readable formats.
In finance, precision is essential. OCR systems are used to process bank statements, transaction records, and compliance documentation. The ability to maintain structural integrity while extracting numerical and textual data is a key requirement that Mistral OCR 4 is designed to support.
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
Document parsing and extraction: Mistral OCR 4 is documented as suited to extracting structure and content from complex, multilingual documents, including PDFs and other document types.
Retrieval-augmented generation: It is documented as suited to producing structured, classified, citation-ready content for semantic chunking and source-grounded answers in RAG pipelines.
Agentic workflows: It is documented as suited to workflows where agents need document structure to complete tasks such as form filling, invoice processing, and compliance checks.
Enterprise search and knowledge bases: It is documented as suited to turning documents into an OCR data source for custom ingestion, entity extraction, and searchable knowledge bases.
import requests headers = {"Authorization": "Bearer " + AIMLAPI_KEY} res = requests.post( "https://api.aimlapi.com/v1/ocr", headers=headers, json={ "model": "mistral/mistral-ocr-4", "document": "https://example.com/invoice.pdf" }, ).json() print(res)
const res = await fetch("https://api.aimlapi.com/v1/ocr", { method: "POST", headers: { Authorization: `Bearer ${AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ model: "mistral/mistral-ocr-4", document: "https://example.com/invoice.pdf", }), }).then((r) => r.json()); console.log(res);
curl -s https://api.aimlapi.com/v1/ocr \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "mistral/mistral-ocr-4", "document": "https://example.com/invoice.pdf" }'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Output | |
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
Mistral OCR 4 This page | Document text extraction | |||
| Document text extraction | ||||
| Document text extraction | ||||
| Document text extraction |
Mistral OCR 4 takes document, image as input and returns text.
Mistral OCR 4 is priced at $5.2 / 1K pages.
Mistral OCR 4 is billed per generation — a fixed charge per output rather than by prompt length.
Mistral OCR 4 was built by Mistral AI.
Use mistral/mistral-ocr-4 as the model id on AI/ML API.
Yes. Mistral OCR 4 is served through AI/ML API, so the same key and endpoint format used for other models applies.
Yes. It adds native paragraph-level bounding box extraction along with structural block labels.
It processes PDFs and images, extracting text, tables, and images from them.
Yes, it is fully backward compatible with Mistral OCR 3.
