Mistral Studio Release
Mistral turned Studio into a platform for building and running AI agents on Nov 5, 2026, with managed agents, internal apps and workflows, mostly in public preview, that run in Mistral's cloud, a company's own cloud or on its own servers. Mistral publishes no Studio price, and evaluations and observability are for Enterprise organizations only.
On Nov 5, 2026 Mistral turned Mistral Studio, its developer platform, into a place for companies to build AI agents and run them in production: managed agents that wake up on triggers, internal apps, and workflows that decide what runs when and where a person signs off. Most of it is a public preview (release notes), and it runs in Mistral's cloud, the customer's own cloud or on its own servers. Mistral's pricing page has no Studio section as of Nov 8, 2026, so there is no published price. In Mistral's launch video, the clip names the three building blocks and how Mistral uses them internally, for incident investigation, product feedback triage and staffing.
What shipped on Nov 5
From Mistral's release notes, twelve entries dated Nov 5, 2026:
| Status | |
|---|---|
| Managed agents, with a Python Agents SDK, background runs that recover from crashes, and human approval | Public preview |
| Workflow builder from plain language, inside Vibe | Public preview |
Mistral Apps command line (mistral apps) |
Public preview |
| Offline and online evaluations, managed datasets, datasets from production traces | Public preview |
| Managed search index | Public preview |
| Webhooks and triggers, with Notion and Lucca ingestion | Public preview |
| Connectors to any HTTP API | Private preview |
| Studio apps shown in Vibe Code Web | Generally available |
The announcement adds a Mistral MCP server, so a coding agent such as Vibe can build and deploy to Studio with one command.
Limits and who can use it
- Evaluations and observability are "available to Enterprise-tier organizations only" (Studio docs).
- Agent sessions end after 24 hours, or after 4 hours with no turn running, and an approval waits one hour by default; all three can be raised (managed agents).
- Managed deployments need a paid plan during the preview (release notes).
- Models: Studio runs Mistral's models, among them Mistral's flagship Mistral Large 4, whose weights Mistral opened on Oct 29, 2026 at $0.70 and $2.10 per million input and output tokens on the API.
- Regions: not stated.
What we think
We think Mistral puts a price on Studio's managed agents by Apr 1, 2027, per run or per hour, and keeps the self-hosted version on Enterprise contracts. A product in public preview with "a paid plan" required already but no rate on the pricing page is a price waiting to be written.
Confidence: Likely. Check by: Mar 2027.
We expect Studio to be Mistral's answer to Anthropic's Managed Agents and OpenAI's Agents API in European tenders: Mistral's own pitch in the video is sovereignty, where it runs, what it reaches and what it learns, which is the one thing its US rivals cannot match on their own clouds.
Confidence: Possible. Check by: Dec 2027.
Sources
- Le Chonk brings the brains. Studio puts it to work., Mistral AI, Nov 5, 2026
- Release notes, Mistral docs, read Nov 8, 2026
- Studio and Managed agents, Mistral docs, read Nov 8, 2026
- Pricing, Mistral AI, read Nov 8, 2026
- Mistral Studio: Agents, apps and workflows for enterprise AI you control, Mistral on YouTube, Nov 5, 2026
Moments
Mistral Studio: Agents, apps and workflows for enterprise AI you control
Mistral
See also
-
MistralMistral AI is a Paris company that makes open-weight and commercial AI models; its flagship, Mistral Large 4, is a 1-trillion-parameter open-weight model released in Oct 2026. It also makes the Vibe assistant (formerly Le Chat) and raised €3 billion in Sep 2026. - Gemini 4 Carbon ReleaseGoogle announced Gemini 4 Carbon on Oct 28, 2026 as its strongest model, priced at $2 and $10 per million input and output tokens for an introductory period, then $4 and $20. As of Nov 8, 2026 only a subset of cyber defenders in Google's Fairwind Program can use it.
- Gemini Omni 1.2 Flash ReleaseGoogle released Gemini Omni 1.2 Flash on Nov 5, 2026, generally available in the Gemini API, Google Cloud, Flow and Google Vids. It makes and edits video clips of 3 to 15 seconds with sound, at about $0.07 a second for 720p, 30% below Omni 1.1, with no free tier.
-
Genie 4 ReleaseGoogle DeepMind announced Genie 4 on Oct 27, 2026, a world model that generates explorable scenes in real time with their own sound and changes them on a typed prompt mid-run. As of Nov 8, 2026 it is a research preview for trusted testers, with Google AI Ultra and the Gemini API promised as "coming soon".
-
GitHub Copilot Fusion ReleaseAt GitHub Universe on Oct 28, 2026, GitHub made Copilot Fusion generally available, letting Copilot route and combine models at three effort levels instead of a model picker, with model costs 30% off through Nov 30, 2026. The same day GitHub Enterprise began including Copilot, and Copilot Standalone went on sale at $10 a user a month.
-
Muse Spaces ReleaseMeta added Spaces to its Muse agent app on Nov 6, 2026, everywhere Muse is available: custom planners, trackers and dashboards you describe in words, which Muse builds and keeps current from your connected apps. Meta lists no price for it; Muse is free for most needs, with paid plans for heavier use.