Mistral AI Launches OCR4 Model Supporting 170 Languages

iconMetaEra
Share
AI summary iconSummary
The French AI firm Mistral AI launched OCR4, a document recognition model supporting 170 languages. It achieved a score of 93.07 on OmniDocBench, surpassing GPT-5.5 Pro and Gemini 3.1 Pro Preview. The model offers enterprise features such as text recognition, border detection, and region classification, with support for RAG semantic chunking. Priced at $4 per 1,000 pages, batch processing receives a 50% discount. This move could test a new resistance level in the AI market, as investor sentiment remains cautious according to the fear and greed index.
Mistral AI has launched its new document recognition model, OCR4, supporting 170 languages across 10 language families and achieving a score of 93.07 on the OmniDocBench benchmark, outperforming GPT-5.5 Pro and Gemini 3.1 Pro Preview. The model offers enterprise-grade features including precise text recognition, bounding box detection, and region classification, and supports RAG semantic chunking. The base API pricing is $4 per thousand pages, designed to make document automation technology more accessible.

Author and source: AIBase

Mistral AI releases the new document recognition model OCR4

French AI startup Mistral AI recently announced the official launch of its proprietary new document recognition model, OCR4. The model excels in processing multilingual documents, supporting 170 languages across 10 language families, offering a new technical solution for global document processing.

Based on publicly available evaluation data, Mistral OCR4 scored 93.07 on the OmniDocBench test, with its output accuracy and natural fluency receiving widespread acclaim, outperforming industry-leading counterparts such as GPT-5.5 Pro and Gemini 3.1 Pro Preview in user experience assessments.

image.png

Mistral AI stated in its release that OCR4 is a model designed for efficient processing. It not only accurately recognizes and extracts text from documents but also includes professional features such as bounding box detection, region classification, and confidence scoring. Additionally, the model has been designed with downstream business needs in mind, supporting RAG (Retrieval-Augmented Generation) semantic chunking, as well as structured output units for agents and connectors, effectively enabling a wide range of complex document intelligence workloads.

For pricing, the Mistral OCR4 Base API is charged at $4 per thousand pages, with a 50% discount available for enterprises opting for batch processing mode; for more advanced document AI services, the pricing is set at $5 per thousand pages. This flexible pricing strategy aims to lower the entry barrier for enterprise users and accelerate the adoption of document automation technologies.

Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.