From the archive
Mistral introduces OCR for PDFs and document images
A document API extracts text, tables, equations, and images for search and other downstream applications.
Mistral introduced an optical character recognition (OCR) service for PDFs and images. The service extracts document content while retaining information about elements such as tables, equations, and embedded images.
Documents become application inputs
A plain text extraction can lose the relationship between a table, its labels, and nearby text. Mistral's service returns ordered text and images, giving document-search and analysis applications more of the original structure to work with. It also supports extracting specific information into structured output.
The release became the document-understanding model in Le Chat and was offered through Mistral's developer platform.
Pricing by the page
The announced API rate was $1 per 1,000 pages, with roughly twice as many pages per dollar through batch processing. That makes document volume, rather than a chat subscription, the starting point for estimating API costs.
The Mistral AI profile covers its hosted services and current access options.