Home
Mistral AI Unveils OCR4: A Multilingual Model Supporting 170 Languages for More Natural Interaction
Mistral AI, a prominent French artificial intelligence innovator, has unveiled OCR4, its latest document recognition model. Engineered to handle multilingual content with ease, OCR4 supports 170 languages spanning 10 distinct language families, offering a robust technical solution for global document processing.
Public evaluation data shows Mistral OCR4 scoring 93.07 on the OmniDocBench benchmark. Praised for its accuracy and natural fluency, it outperforms industry leaders like GPT5.5Pro and Gemini3.1Pro Preview in user experience assessments.

According to the release, OCR4 is optimized for efficiency. Beyond precise text recognition and extraction, it offers advanced features such as border detection, area classification, and confidence scoring. Designed with downstream applications in mind, the model supports RAG (retrieval-augmented generation) semantic chunking and provides structured content output via connectors, effectively handling complex intelligent document workflows.
Regarding pricing, the base API for Mistral OCR4 is available at $4 per thousand pages. Companies opting for batch processing receive a 50% discount, while premium document AI services are priced at $5 per thousand pages. This flexible pricing structure aims to reduce entry barriers for enterprise users and accelerate the adoption of document automation technology.
Related article
Google Debuts Gemini Omni 1.1 Flash Video Model in Direct 4K Resolution
The race for large-scale multimodal models is accelerating, with a clear shift toward higher resolutions and extended video generation capabilities. On August 27, Google unveiled the Gemini 1.1 Flash video generation model, a significant upgrade that
Amazon Wins Court Injunction Against PerplexityAI, Blocking Shopping and Mandating Data Deletion
On March 10, 2026, the U.S. District Court for the Northern District of California issued an injunction against Amazon, restricting the AI browser proxy Comet used by the AI startup Perplexity to perform shopping tasks on Amazon's platform. This ruli
OpenAI Slows Rollout to Address Security Holes and Refine AI Models
Sam Altman, OpenAI’s CEO. Photo: Chip Somodevilla/Getty ImagesAfter the Hugging Face security breach, OpenAI has slowed its development pace. CEO Sam Altman emphasizes that alignment across all training phases is critical.OpenAI is taking a step back
Related Special Topic Recommendations
Comments (0)
0/500
Mistral AI, a prominent French artificial intelligence innovator, has unveiled OCR4, its latest document recognition model. Engineered to handle multilingual content with ease, OCR4 supports 170 languages spanning 10 distinct language families, offering a robust technical solution for global document processing.
Public evaluation data shows Mistral OCR4 scoring 93.07 on the OmniDocBench benchmark. Praised for its accuracy and natural fluency, it outperforms industry leaders like GPT5.5Pro and Gemini3.1Pro Preview in user experience assessments.

According to the release, OCR4 is optimized for efficiency. Beyond precise text recognition and extraction, it offers advanced features such as border detection, area classification, and confidence scoring. Designed with downstream applications in mind, the model supports RAG (retrieval-augmented generation) semantic chunking and provides structured content output via connectors, effectively handling complex intelligent document workflows.
Regarding pricing, the base API for Mistral OCR4 is available at $4 per thousand pages. Companies opting for batch processing receive a 50% discount, while premium document AI services are priced at $5 per thousand pages. This flexible pricing structure aims to reduce entry barriers for enterprise users and accelerate the adoption of document automation technology.
Google Debuts Gemini Omni 1.1 Flash Video Model in Direct 4K Resolution
The race for large-scale multimodal models is accelerating, with a clear shift toward higher resolutions and extended video generation capabilities. On August 27, Google unveiled the Gemini 1.1 Flash video generation model, a significant upgrade that
Amazon Wins Court Injunction Against PerplexityAI, Blocking Shopping and Mandating Data Deletion
On March 10, 2026, the U.S. District Court for the Northern District of California issued an injunction against Amazon, restricting the AI browser proxy Comet used by the AI startup Perplexity to perform shopping tasks on Amazon's platform. This ruli
OpenAI Slows Rollout to Address Security Holes and Refine AI Models
Sam Altman, OpenAI’s CEO. Photo: Chip Somodevilla/Getty ImagesAfter the Hugging Face security breach, OpenAI has slowed its development pace. CEO Sam Altman emphasizes that alignment across all training phases is critical.OpenAI is taking a step back











