Mistral AI

Mistral AI unveiled Mistral Large 2 (123B parameters), achieving 84.0% MMLU and rivaling GPT-4o in multilingual coding and reasoning at half the inference price.

Mistral AI has established a new research hub in Munich, Germany. The facility is dedicated to the development of Physics AI and Industrial AI applications.

Mistral AI and Mozilla have formed a partnership to integrate Mistral's open-weight models directly into the Firefox browser ecosystem. This collaboration enables local, private execution of multilingual AI models within the user's web browser environment.

Cloudera has integrated Mistral AI's open-weight models into the Cloudera Machine Learning (CML) platform. This integration enables enterprise users to deploy Mistral models directly within their private cloud environments for data-sovereign AI workloads.

Mistral AI has demonstrated a specialized workflow for automated code migration from Fortran 77 to C++. This capability utilizes AI agents to refactor 40,000 lines of legacy code into modern programming languages.

Mistral AI has secured €3 billion in Series D funding. This capital injection brings the company's post-money valuation to over €21 billion.

Mistral AI has entered into a strategic partnership with HUMAIN to integrate Mistral's generative AI models into HUMAIN's specialized enterprise platforms. This collaboration enables the deployment of Mistral's large language models within HUMAIN's existing infrastructure for targeted industry applications.

Mistral AI has introduced 'Agentic Search', a specialized retrieval layer designed for document navigation and information verification. This feature enables AI systems to autonomously read and cross-reference data within complex document structures.

Mistral AI is deploying dedicated European-based inference infrastructure to support sovereign AI requirements. The company is expanding its portfolio of open-weight models to provide localized control over AI workloads.

Shieldstral introduces a 3B open-weights multimodal safety classifier that outperforms models up to 7x its size....

Studio provides a system of record for AI prompts and skills—versioned, owned, and traceable. Iterate fast, ship with control, and ensure consistent AI behavior....

Introducing Robostral Navigate: 8B model achieving 76.6% on R2R-CE with just a single RGB camera. No depth sensors, LiDAR, or multiple cameras needed....