What happened
Mistral AI shipped OCR 4, its latest optical character recognition model, featuring bounding boxes, block classification, and inline confidence scores alongside extracted text. The model supports 170 languages across 10 groups and deploys in a single container for fully self-hosted enterprise use. OCR 4 achieved an 85.20 score on OlmOCRBench and a 72% human preference win rate over tested competitors. It integrates with Mistral Search Toolkit and offers API pricing at $4 per 1,000 pages, with a $2 batch discount.
Why it matters
Organisations requiring secure, high-throughput document processing gain a new self-hosted option for data residency and compliance. OCR 4's structured output, including bounding boxes and block types, provides citation-ready inputs for RAG and enterprise search, reducing downstream processing costs. An AI engineer at Rogo reported 8x lower cost and 17x lower latency compared to leading agentic document parsers for financial QA. This follows Mistral AI's recent launch of its full AI stack, expanding its enterprise offerings.
