Mistral AI

Mistral Large 4 is Mistral AI's flagship multimodal large language model capable of extracting and understanding text from images, including complex layouts, handwriting, and scene text across multiple languages. It stands out for its strong multilingual capabilities and reasoning-driven OCR that interprets context beyond simple character recognition.

mistral-ocr-2505 — OCR-capable vision model.

Mistral dedicated OCR endpoint with page-wise Markdown and optional structured block metadata.

200 pages / $1$0.005/page

mistral-ocr-2503 — OCR-capable vision model.

labs-devstral-small-2512 — OCR-capable vision model.

~3,850 pages / $1*~$0.00026 per standard page

Pixtral Large — Mistral's vision-language model. Strong document understanding with 128K context.

~192 pages / $1*~$0.0052 per standard page

magistral-medium-2509 — OCR-capable vision model.

~334 pages / $1$0.002996/page