Introducing Mistral OCR 4

Mistral AI released Mistral OCR 4, a document parsing model that returns bounding boxes, typed-block classification and inline confidence scores alongside extracted text. The company said the model supports 170 languages across 10 language groups and runs in a single container for fully self-hosted deployments.

OCR 4 accepts PDF, DOC, PPT and OpenDocument formats. Bounding boxes localize text for in-context highlighting and data pipelines, while block types and confidence scores support source-grounded citations, redactions and human-in-the-loop verification, according to Mistral. The company described OCR 4 as a compact model suited to cost-sensitive and high-volume deployments, with self-managed deployment available to enterprise customers.

Pricing is $4 per 1,000 pages through the API, with a 50% Batch-API discount reducing it to $2 per 1,000 pages. Document AI is priced at $5 per 1,000 pages. The model is available via API through Mistral Studio, Amazon SageMaker and Microsoft Foundry, with Snowflake Parse Document listed as coming soon.

Mistral said independent annotators preferred OCR 4 over every leading OCR and document-AI system tested, with win rates averaging 72% across 600+ documents in 12+ languages. OCR 4 scored 85.20 on OlmOCRBench and 93.07 on OmniDocBench, and led the company's internal Crawl Multilingual evaluation at .98. Mistral said both public benchmarks have known scoring limitations, including ground-truth errors, equivalent math notation, equation segmentation, multi-column reading order and block-type attribution, and described the aggregate scores as directional rather than definitive. Competitor scores reflect internal reproductions.

OCR 4 is an ingestion component of the Mistral Search Toolkit, announced at the AI Now Summit and in public preview. Mistral said the model is not intended for medical diagnosis, legal advice or judgment, high-stakes financial decisions, safety-critical systems, real-time processing or non-document inputs. A webinar is scheduled for July 7th at 6:00 PM CET.