On 6 March 2025, Mistral AI unveiled Mistral OCR, an advanced document interpretation API that offers processing of 1,000 pages for $1, surpassing competitors with an impressive accuracy rate of 94.89%.
The service is capable of interpreting complex documents, including images, tables, and mathematical formulas, while preserving the document’s structure and hierarchy. The technology is particularly valuable when combined with RAG (Retrieval Augmented Generation) systems that process multimodal documents. Mistral OCR stands out from competitors with its performance—according to official benchmark tests, it achieves an overall accuracy of 94.89%, compared to Google Document AI’s 83.42%, Azure OCR’s 89.52%, and GPT-4o’s 89.77%. The service offers multilingual capabilities, supporting thousands of scripts, fonts, and languages with a 99.02% match rate in generation. Its processing speed is especially noteworthy, capable of handling up to 2,000 pages per minute.
The service offers a diverse range of applications, including the digitisation of academic research, preservation of historical and cultural heritage, and optimisation of customer service. The Mistral OCR API is available for a free trial on the Le Chat platform and is accessible via the La Plateforme developer interface, with availability soon to extend to cloud and inference partners as well as on-premises deployment. Document files must not exceed 50 MB in size or 1,000 pages in length, while the service remains competitively priced at just $0.001 per page. Mistral OCR is already widely available to business users and has made a significant impact across various industries, particularly in the financial, legal, and healthcare sectors, where accurate and rapid document processing is critical.
Sources:
1.
2.
3.
