Mistral AI released OCR 4.1, its latest optical character recognition service that powers the company's Document AI stack with enhanced extraction capabilities.

The new service introduces native paragraph-level bounding box extraction, structural block labels, and block-level confidence scores. These features represent a significant upgrade from previous OCR offerings in the market.

Pricing and availability

Mistral priced OCR 4.1 at $4 per 1,000 pages for basic processing, with annotated pages costing $5 per 1,000. The service launched on July 16, 2026, and is available through Mistral's API endpoints.

The company positioned OCR 4.1 as a "premier" service within its broader document processing suite. Users can access the tool through Mistral's playground interface or integrate it directly via API calls.

Technical capabilities

OCR 4.1 supports multiple processing modes including basic OCR extraction, bounding box detection, and structured annotations. The service can handle batch processing through Mistral's /v1/batch endpoint.

The paragraph-level extraction feature allows users to identify document structure automatically. Block-level confidence scores help developers assess the reliability of extracted text segments.

Structural block labels categorize different document elements, enabling more sophisticated document understanding workflows. This positions the service for enterprise document processing applications.

Market context

The OCR market has seen increased competition as companies integrate document processing into AI workflows. Major cloud providers and specialized AI companies have launched similar services targeting enterprise customers.

Mistral's entry builds on its foundation model expertise, applying its language understanding capabilities to document processing tasks. The company has expanded beyond pure language models into multimodal applications.

OCR 4.1 joins Mistral's broader model lineup, which includes the Z.ai GLM series and Shieldstral safety models. The company continues developing specialized AI services for enterprise use cases.