Mistral launches Vibe consumer agent and Industrial Engineering stack on new 128B Medium 3.5

Mistral AI unveiled three coordinated announcements on May 28. First, Mistral Medium 3.5 — a new flagship 128B dense (not MoE) model for instruction-following, reasoning, and coding, capable of self-hosting on as few as four GPUs, a key differentiator for enterprises that don't want to depend on hyperscaler APIs. Second, the consumer assistant was rebranded to 'Mistral Vibe' and gained two new modes: Work Mode for long-horizon productivity tasks and Code Mode for remote coding agents, both powered by Medium 3.5. Third, 'Mistral for Industrial Engineering' bundles Mistral LLMs with physics simulation from the Emmi AI acquisition — targeting manufacturing customers with a stack that can reason about CAD, simulate, and generate engineering plans.

On infrastructure, Mistral announced a new 10MW inference facility in Les Ulis to secure European compute capacity, building on its sovereignty-positioning. And Microsoft Copilot Studio added Mistral Medium 3.5 to its model lineup, citing in-region data control and strong multilingual performance — a notable deal given Microsoft is simultaneously building in-house coding models to reduce dependence on third parties.

Mistral's official X account confirmed the launch: 'SHIPPED. Mistral Vibe is now the AI agent for long-horizon productivity and coding, and the home for Work mode, Code mode...' The Mistral AI Now Summit notes circulated widely on Hacker News (314 points), reigniting debate about Europe's ability to compete with US frontier labs on compute. Skeptical takes: Vibe is Mistral's third consumer-product rebrand in two years, and Hacker News commenters questioned whether Industrial Engineering is a serious vertical play or a marketing wrapper.

Watch next: Medium 3.5 benchmark numbers versus Llama, DeepSeek V4, and Qwen3.7-Max; whether the 4-GPU self-host claim holds at production latency; and whether Emmi-powered industrial customers materialize beyond launch partners.