---
title: "Mistral AI Studio Adds Agents in Public Preview"
locale: en
category: news
category_name: "News"
translation_status: reviewed
license: cc_by
author: "Injoys Editorial Team"
source_url: https://injoys.com/en/articles/mistral-ai-studio-expands-managed-agents-public-preview
published_at: 2026-11-06T01:30:59+09:00
---

# Mistral AI Studio Adds Agents in Public Preview

> Mistral AI added managed agents and apps to Studio and expanded workflow, tracing, and evaluation features. Agents and workflows are in public preview; availability in Korea and separate execution fees have not been confirmed.

## Key Points

- On November 5, 2026, Mistral AI announced the addition of managed agents and apps to Studio and the expansion of related features.
- Managed Agents, Agents SDK, and Mistral Workflows are in public preview and remain subject to change.
- Apps cannot be created on Free or on Pro and Team plans without pay-as-you-go billing enabled.
- Mistral Large 4 API pricing is $0.7 per 1 million input tokens and $2.1 per 1 million output tokens.
- Sign-up, payment, and service availability in Korea, as well as support for data storage in Korea, have not been confirmed.

Mistral AI announced an expansion of Studio on November 5, 2026. It added managed agents and apps. Agents are in public preview, and app creation is restricted by plan. Availability in Korea was not specified.

This article follows Mistral AI's announcement and official documentation, checked as of November 6, 2026.

## Managed agents and apps added to Studio

Mistral AI said it had expanded Studio's development and deployment features. The announcement introduced managed agents and apps. Workflow and enterprise context features were also part of the expansion. Agent observability and evaluation features were included as well.

Le Chonk is the name of the Mistral Large 4 model. The company introduced it as a native multimodal model in public preview. Studio serves as the environment for developing and running agents and apps.

The official release notes explained how the workflow builder works. It turns tasks described in natural language into working workflows. Vibe handles authoring, testing, and deployment.

## Agents and workflows in public preview, with feature-specific access conditions

Managed Agents and Agents SDK are in public preview. Mistral Workflows is also available in public preview. The official documentation separately specifies possible changes and access conditions for each product.

| Feature | Release status or access conditions | Official documentation's description |
| --- | --- | --- |
| Managed Agents and Agents SDK | Public preview | APIs, behavior, and limits may change |
| Mistral Workflows | Public preview | No major changes are planned, but APIs and features may still change |
| Agent and workflow execution | Mistral cloud or your own infrastructure | Both execution environments are supported |
| Tracing and evaluation | Enterprise-tier organizations only | Restrictions apply based on organization tier |
| App creation | Restricted by plan and whether pay-as-you-go billing is enabled | Unavailable on Free or on Pro and Team without pay-as-you-go billing enabled |

According to the Studio documentation, tracing and evaluation are available only to **Enterprise-tier organizations**. The conditions for running agents must be distinguished from those for using tracing and evaluation. Whether users can sign up or use the service in Korea was not specified.

## Model API pricing and app configuration limits disclosed

The official API pricing table lists separate input and output rates for Mistral Large 4. The prices below were checked as of November 6, 2026. Separate rates for running managed agents, workflows, and apps were not confirmed.

| Billing item | Unit | Official listed price |
| --- | --- | --- |
| Mistral Large 4 input | 1 million tokens | $0.7 |
| Mistral Large 4 output | 1 million tokens | $2.1 |

These prices alone cannot be used to calculate the total cost of operating Studio. Model API charges and managed execution costs must be checked separately. The fact that separate execution rates were not confirmed does not mean execution is free.

A single app can have **at most 3 frontends and backends combined**. The official managed app documentation excludes workers from this limit. This does not mean that 3 frontends and 3 backends are allowed.

On Free, users cannot create apps even if deployment commands are visible. The same restriction applies to Pro and Team without pay-as-you-go billing enabled. App deployment permissions should not be judged solely by whether commands are displayed.

## Model size figures are self-reported by Mistral AI

The size figures for Mistral Large 4 are specifications listed in the company's model card. Active parameters and total parameters are separate items. The figures below are not performance measurements from an independent organization.

| Model card item | Listed figure | Korean quantity notation |
| --- | --- | --- |
| Active parameters | 52B | 52 billion |
| Total parameters | 1.05T | 1 trillion 50 billion |
| Vision encoder | 1.6B | 1.6 billion |

Mistral AI's model introduction describes Le Chonk as a 1 trillion-parameter model. The model card gives a more specific total parameter count of 1.05T. These numbers alone cannot establish accuracy on Korean-language tasks or execution speed.

## Korea support, execution pricing, and general availability schedule remain unconfirmed

The official sources checked did not specify some conditions needed for adoption in Korea. The end date for the public preview was also not confirmed. Public preview should not be treated as general availability.

- Availability of sign-up, payment, and service in Korea
- Support for a Korean region and data storage within Korea
- Separate billing rates for running managed agents, workflows, and apps
- Schedule for the end of public preview and general availability
- Exact announcement time and time zone for Studio's new features

The publication date associated with the announcement is November 5, 2026. The exact announcement time and the company's reference time zone were not confirmed. The announcement time converted to Korean time therefore cannot be established either.

## Check account, cost, and data conditions before adoption in Korea

Before adopting the service in Korea, first check whether it is available there. Support for running on your own infrastructure does not, by itself, guarantee data storage within Korea. Before paying, it is advisable to check the following items in the official documentation and with the company.

- Whether Korean accounts can sign up, pay, and use the service
- App creation permissions and conditions for enabling pay-as-you-go billing on the intended plan
- Whether the Enterprise tier required for tracing and evaluation applies
- Execution costs incurred in addition to model API charges
- Data storage locations for each execution environment and support for a Korean region
- Whether you can respond to changes in APIs, behavior, and limits during preview

## FAQ

### Can the new Studio features be used right away in Korea as well?
Official sources do not specify whether sign-up, payment, and service are available in Korea. Support for a Korea region and data storage within Korea has also not been confirmed, so you should check with the company before adopting it.

### Can you build and deploy apps on the Free plan?
According to the Studio app documentation, you cannot create apps on Free even if you can see deployment commands. The same restriction applies to Pro and Team if pay-as-you-go billing is not enabled.

### Can operating costs be calculated using only the Mistral Large 4 API pricing?
Input costs $0.7 per 1 million tokens, and output costs $2.1 per 1 million tokens. Separate rates for running managed agents, workflows, and apps have not been confirmed, so this pricing alone cannot be used to calculate total operating costs.

## Sources

- [Le Chonk brings the brains. Studio puts it to work. - Mistral AI](https://mistral.ai/news/build-with-studio/)
- [Mistral Large 4 (Le Chonk) is now open weight - Mistral AI](https://mistral.ai/news/mistral-large-4-oss/)
- [Managed Agents | Mistral Docs - Mistral AI](https://docs.mistral.ai/studio/agents/overview)
- [Release notes | Mistral Docs - Mistral AI](https://docs.mistral.ai/resources/release-notes)
- [Overview | Mistral Docs - Mistral AI](https://docs.mistral.ai/studio/workflows/overview)
- [Studio | Mistral Docs - Mistral AI](https://docs.mistral.ai/studio)
- [Apps overview | Mistral Docs - Mistral AI](https://docs.mistral.ai/studio/apps/overview)
- [Deploy apps with managed deployments | Mistral Docs - Mistral AI](https://docs.mistral.ai/studio/deployments/managed/apps)
- [API pricing | Mistral Docs - Mistral AI](https://docs.mistral.ai/inference/pricing)
- [Mistral Large 4 - Mistral AI | Mistral Docs - Mistral AI](https://docs.mistral.ai/models/mistral-large-4-0)

## Images

![A woman engineer in blue workwear inspects a server pulled out from a rack in a data center.](https://injoys.com/rails/active_storage/blobs/proxy/eyJfcmFpbHMiOnsiZGF0YSI6Mzg0NjQsInB1ciI6ImJsb2JfaWQifX0=--69bd002e677addff6121ae219d07e6df4704afff/ai-034cade6.webp)
![A man examines a workflow on a tablet at a testing bench, with mobile devices on shelves behind him.](https://injoys.com/rails/active_storage/blobs/proxy/eyJfcmFpbHMiOnsiZGF0YSI6Mzg0NjcsInB1ciI6ImJsb2JfaWQifX0=--06d7b31138a271f91f7d8b467f2fc0079fb5d258/ai-1c094ba6.webp)