Mistral Buys SkipLabs, Its Fifth Deal of 2026, as Studio Adds Managed Agents

Mistral AI, the Paris-based model maker, made two moves on Thursday, Nov. 5, aimed at large companies. It expanded Mistral Studio, its enterprise platform, with managed agents and hosted apps. It also bought SkipLabs, a Paris- and London-based startup that builds AI code-generation technology, Maddyness reported. SkipLabs is Mistral’s fifth acquisition of 2026.

The two announcements share a goal: making Mistral’s flagship open-weight model, Mistral Large 4, nicknamed “Le Chonk,” useful inside banks, manufacturers and governments that want control over where their data and AI run. Mistral’s own tagline: “Le Chonk brings the brains. Studio puts it to work.”

What’s new in Mistral Studio

Mistral’s announcement adds managed agents and apps alongside new capabilities for workflows, enterprise context, agent observability and evaluations. According to Mistral’s blog and changelog, as summarized by Neural Brew:

• Managed agents (public preview) can carry “skills” that describe team procedures, run code in sandboxes, use sub-agents and add custom logic through hooks.
• Workflows (now generally available) mix fixed rules, agent reasoning and human approvals, resume where they stopped, and can be triggered by events in company systems.
• Apps (public preview) come with built-in hosting, sign-in and access controls.
• A new Mistral MCP server lets developers build with Mistral Vibe or another coding agent, test locally and deploy to Studio in one command, with runs traced in the OpenTelemetry format.

All three features are live on Mistral’s cloud now. Mistral says they will reach private-cloud and on-premises deployments in late November. Neural Brew noted that no pricing for managed agents, apps or workflows had been published.

Key facts: the deals and the customers

The deal terms for SkipLabs were not disclosed in the report. Verlaguet spent nearly nine years at Meta, including at its fundamental AI research lab, Maddyness said.

The model behind Studio

Studio runs on Mistral Large 4, whose weights Mistral released on Hugging Face earlier this month. IT Brief reported that the model has 1 trillion parameters, 52 billion of them active on any given task, and that it is natively multimodal. Mistral said it trained the model for three more weeks with reinforcement learning after an earlier preview. It also said Large 4 improved its Terminal-Bench 4.0 score from 28% to 41% and uses 47% fewer tokens per task than China’s GLM-5.3 for similar results on Artificial Analysis’ coding measure. Those are company claims. Mistral is pitching the open weights to organizations that want to run coding and cyber-defense agents on their own servers, which is exactly the deployment Studio’s late-November on-premises release targets.

Why control is the selling point

Mistral’s customer quotes focus on keeping AI inside their own walls. “The AI systems we build run some of our most sensitive processes, so we run Mistral Studio on our own infrastructure and keep those systems under our control,” Su Yang, head of artificial intelligence and IT innovation at BNP Paribas, said in Mistral’s post. ABANCA’s Jose Jorge Garcia Romaris said running Studio in the bank’s own cloud, with inference provided by Mistral, lets it bring AI into the business “while retaining control over our infrastructure and how sensitive customer data is accessed and used.”

That pitch matches the company’s recent direction. Beyond chatbots, Mistral has signed industrial and scientific partnerships, including a planned physics model with CERN. Maddyness noted that Chinese labs still lead several open-model rankings, while Mistral says Large 4 stands out in cybersecurity, finance and industrial uses. For more on how the company fits the field, see our Mistral AI profile.

Sources

• Mistral AI — Le Chonk brings the brains. Studio puts it to work., Nov. 5, 2026
• Maddyness — Mistral s’offre une cinquième acquisition depuis le début de l’année, Nov. 5, 2026
• Neural Brew — Mistral Studio Adds Managed Agents and Apps, Deployable On-Premises in Late November, Nov. 5, 2026
• IT Brief — Mistral releases open-weight ML4 for coding & cyber, Nov. 2, 2026