Mistral AI

5 articles tagged with Mistral AI

May 28, 2026
product updateMistral AI

Mistral AI Expands Into Industrial Engineering With Airbus, BMW Partnerships and Acquires Physics AI Firm Emmi

Mistral AI announced a new industrial engineering AI stack combining physics models with partnerships across aerospace, automotive, and semiconductor sectors. The company acquired scientific AI firm Emmi on May 22, 2026, and is opening a 10 MW inference data center in Les Ulis, France in Q3 2026.

model releaseMistral AI

Mistral Releases Medium 3.5: 128B Model with Cloud Coding Agents and 77.6% SWE-Bench Verified

Mistral AI released Medium 3.5, a 128B dense model with a 256k context window that scores 77.6% on SWE-Bench Verified. The model powers new remote coding agents in Mistral Vibe that run asynchronously in the cloud, plus a new Work mode in Le Chat for multi-step agentic tasks.

product updateMistral AI

Mistral AI Releases MCP Connectors in Studio with Direct Tool Calling and Human-in-the-Loop Workflows

Mistral AI has released Connectors in Studio, allowing developers to integrate custom MCP (Model Context Protocol) servers alongside built-in connectors for enterprise AI applications. The release includes direct tool calling, human-in-the-loop approval flows, and programmatic connector management via API and SDK.

product updateMistral AI

Mistral AI launches Forge, enterprise platform for training custom models on proprietary data

Mistral AI has launched Forge, a platform for enterprises to train custom AI models on proprietary data including codebases, compliance policies, and operational records. Early partners include ASML, DSO National Laboratories Singapore, Ericsson, European Space Agency, and HTX Singapore.

researchMistral AI

Mistral AI fine-tunes Pixtral-12B on satellite imagery, boosting classification accuracy from 56% to 91%

Mistral AI reports that fine-tuning its Pixtral-12B vision model on satellite imagery increased classification accuracy from 56% to 91% on the Aerial Image Dataset. The company used LoRA (Low-Rank Adaptation) to train on 8,000 samples for under $10, reducing hallucinations from 5% to 0.1%.