Mistral OCR 4.1
OCR 4.1 is Mistral's document extraction service released in July 2026. It adds paragraph-level bounding boxes, structural block labels and confidence scores for individual blocks. These outputs retain information about the position and type of content, in addition to the extracted text.
Keep the location of each passage
The model page lists bounding-box extraction, OCR and structured annotations on the OCR endpoint. Batch processing is also available. The paragraph location, block label and confidence are distinct outputs; a text transcript alone does not retain all of them.OCR 4.1 API model page
Build review around the extracted blocks
Practical guidance. A document viewer can use paragraph locations to connect an extracted field to its source page. Confidence can help decide what a person should review, but it should be checked against actual mistakes in the document types being processed. Evaluate merged tables, handwriting and reading order separately.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API |
| Model identifiers | mistral-ocr-4-1 |
| Initial release stage | General availability.OCR 4.1 API model page |
| API | mistral-ocr-4-1 through /v1/ocrOCR 4.1 API model page |
| Access terms | Premier service; public weight download not confirmedOCR 4.1 API model page |
Related models
- Mistral Small 4
Mistral AI · Mistral Small
An Apache-licensed multimodal model that combines instruction following, reasoning and coding with selectable reasoning modes.
Launch date not confirmedAPI · Downloadable - Mistral OCR 3
Mistral AI · Mistral OCR
An OCR model with HTML table output and support for handwritten forms, available through commercial document-processing services.
Launch date API · Provider product