Mistral introduces advanced search tool to improve document accuracy

Mistral AI has launched an enhanced version of its Agentic Search technology, designed to address challenges in retrieving and analyzing complex enterprise documents. Traditional retrieval-augmented generation (RAG) methods have struggled with structured documents and cross-source comparisons due to their inability to reason or iterate. Agentic Search introduces a multi-step retrieval loop mechanism, allowing models to refine search strategies, verify information across multiple documents, and improve retrieval accuracy. In industry benchmark tests, the system has demonstrated significant improvements, achieving an accuracy rate of 86% for complex documents. The technology is built on existing enterprise search indexes and provides models with five core tools similar to file system operations, enabling more flexible and precise data handling. Agentic Search is compatible with both cloud and on-premise deployments, supporting integration into customer workflows through the Mistral Search Toolkit. This advancement aims to help enterprises unlock the value of sensitive domain data and build high-precision retrieval systems.