Mistral OCR 2503
Mistral OCR 2503 introduced the provider’s dedicated document-recognition model. It accepts PDF pages and images and extracts content with the document’s text, tables, equations and illustrations in mind.
Deprecated API; check access details
Document structure beyond plain text
The launch describes ordered, interleaved text and images and shows Markdown output. It targets complex layouts, including scientific papers with tables, formulas and figures. Extracted content can then enter a document-search pipeline without first reducing every page to plain text.Mistral OCR 2503 release announcement
Hosted and negotiated self-hosting routes
The launch offered an API and selective self-hosting for organizations with sensitive documents. That arrangement did not establish a public weight download. The model record now marks the original mistral-ocr-2503 version deprecated.Mistral OCR 2503 release announcementMistral OCR 2503 model and lifecycle record
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Deprecated API; check access details |
| Launch date | |
| Access | API |
| Model identifiers | mistral-ocr-2503 |
| Weight access | No public weight download is listed in the official model record.Mistral OCR 2503 model and lifecycle record |
| Access and lifecycle | Hosted API deprecated on 2 December 2025.Mistral OCR 2503 model and lifecycle record |