Mistral AI, a rising star in the field of artificial intelligence in France, recently announced the launch of its newly developed document recognition model, OCR4. The model excels in handling multilingual documents, supporting 170 languages across 10 language families, providing a new technical solution for global document processing.

According to publicly available evaluation data, Mistral OCR4 achieved a score of 93.07 on the OmniDocBench test. Its output content has been widely praised for its accuracy and natural fluency, performing better than industry-leading products such as GPT5.5Pro and Gemini3.1Pro Preview in user experience evaluations.

image.png

Mistral AI mentioned in the release that OCR4 is a model focused on efficient processing. It not only accurately recognizes and extracts text from documents, but also features professional functions such as border recognition, area classification, and confidence scoring. In addition, the model has been designed with downstream business needs in mind, supporting RAG (retrieval-augmented generation) semantic chunking, as well as basic units of intelligent agents and structured content output through connectors, effectively supporting various complex document intelligent workloads.

In terms of pricing, the base API for Mistral OCR4 is charged at $4 per thousand pages. If companies choose the batch processing mode, they can enjoy a 50% price discount; for more professional document artificial intelligence services, the price is set at $5 per thousand pages. This flexible pricing strategy aims to lower the usage threshold for enterprise users and accelerate the popularization of document automation technology.