Mistral OCR 4 est sorti !
Bounding boxes, détection intelligente (tableaux, équations, titres), scores de confiance et 170 langues
72 % de préférence humaine • 85,2 sur OlmOCRBench
https://mistral.ai/news/ocr-4/

Mistral released OCR 4, a document model built for RAG. Instead of dumping text it returns structure: bounding boxes, typed blocks (tables, equations, signatures), and confidence scores, for source-grounded citations and human review. 170 languages, self-hostable in one container.

Mistral OCR 4 est sorti : bounding boxes, classification de blocs, scores de confiance, 170 langues, déployable en self-hosted dans un seul container. Le tout à $2/1000 pages en batch. ⬇️
https://news.humancoders.com/t/machine-learning/items/59129-mistral-ocr-4-reconnaissance-de-documents-avec-bou

📝 Mistral OCR 4 vs AWS Textract vs Google Document AI: The Cheapest Accurate Document API (2026)

Mistral shipped OCR 4 on June 23, 2026 — model `mistral-ocr-latest` — and it tops OlmOCRBench at 85.20, handles 170…

https://rohitraj.tech/en/notes/mistral-ocr-4-vs-textract-google-document-ai-2026

Just saw an open‑source OCR model hit 82.4 on the olmOCR‑bench—handles equations, tables, multilingual docs, and scales like a champ with PaddleOCR VL and ERNIE‑4.5‑0.3B. Dive into the details! #OCR #olmOCRbench #VisionLanguage

🔗 aidailypost.com/news/open-so...

This tweet appeared under this Techmeme headline:

@mistralai:

On public benchmarks, OCR 4 tops OlmOCRBench (85.20) and leads our internal multilingual eval, with the widest gains on rare and low-resource languages, where most systems fall off.

Mistral AI launched OCR 4! 📑 It introduces bounding boxes, block classification, and confidence scores for 170 languages. It runs fully self-hosted in a single container and achieved a top 85.20 score on the OlmOCRBench public benchmark.

mistral.ai/news/ocr-4/