Mistral AI Launches OCR 4.1 for Document AI Stack
Pricing set at €3.5 per 1,000 pages as the French AI firm updates its document processing tools
Mistral AI has officially released OCR 4.1, an optical character recognition service designed to support its Document AI stack. The update is now available in public preview, marking a significant step in the company’s efforts to refine its document processing infrastructure for enterprise and developer use.
The latest iteration introduces native paragraph-level bounding box extraction, allowing for more precise identification of text segments within scanned documents. Alongside this feature, the service includes structural block labels and block-level confidence scores, providing users with greater granularity and reliability when parsing complex layouts.
Commercial availability comes with a defined pricing structure. Mistral AI has set the cost at €3.5 per 1,000 pages for standard processing. For users requiring annotated outputs, the rate increases to €4.38 per 1,000 annotated pages, positioning the service within a competitive tier for high-volume document analysis.
The release documentation, published via Hacker News on 13 August 2026, lists the launch date as 16 July 2026. This timeline places the announcement in the future relative to the current operational date, suggesting the source material may reflect a hypothetical scenario or contain a temporal discrepancy in its reporting.
As a public preview, the software remains in a testing phase and is subject to change. Investors and developers monitoring the artificial intelligence sector should note that performance metrics and accuracy rates are not yet detailed in the primary source material, and claims regarding the service’s capabilities rely solely on the company’s own summary.
