Mistral Studio Brings Enterprise Agents to VPC and On-Prem
Mistral Studio adds hybrid deployment options across VPC and on-premises infrastructure, running consistent agentic workflows powered by Mistral Large 4.
Mistral Studio expands its enterprise agent platform to private cloud and on-premises environments at the end of November 2026, so agent runtimes no longer have to run in a public cloud. The architecture allows organizations to build a single agentic workflow and execute it with consistent behavior across hosted cloud instances, private VPCs, or sovereign on-premises servers.123
Workflows are currently generally available on Mistral Cloud, while Apps and Managed Agents operate in public preview before the full hybrid deployment milestone. For teams managing regulated data or legacy systems, this hybrid model addresses infrastructure boundaries that prevent organizations from deploying purely cloud-native setups like those found in AI agent orchestration. Customers already running Mistral Studio in production environments include ABANCA, Accenture, Alten, BNP Paribas, Capgemini, CMA CGM, and Tata Consultancy Services.12
Core components of the studio architecture
Mistral Studio organizes automation into Workflows that define how processes run, establishing which steps follow fixed rules, where an agent reasons, and when a person decides. This setup keeps automated execution predictable while ensuring human verification at defined checkpoints.1
Developers create Managed Agents equipped with dedicated procedures and sandboxes where they run code, delegating parts of a task to subagents. To bridge internal infrastructure, Studio links models to company systems through connectors and Model Context Protocol (MCP) servers.1
Powering reasoning with Mistral Large 4
At the core of the platform is Mistral Large 4, also known as "Le Chonk", engineered for heavy reasoning tasks such as cybersecurity audits and complex coding workflows. Connecting an agent to an MCP server allows Le Chonk to parse structured technical data and make multistep decisions across legacy databases.1
The platform supports a tiered model strategy to balance reasoning capability against operational token costs. Organizations can deploy smaller models for routine intermediate steps while reserving flagship models for final decisions. Detailed tier costs appear on our Mistral API pricing tracker.4
| Model | Input per 1M Tokens | Output per 1M Tokens |
|---|---|---|
| Mistral Large 4 | $0.70 | $2.10 |
| Mistral Large 3 | $0.50 | $1.50 |
| mistral-large-2512 | $0.50 | $1.50 |
| Mistral Medium 3.5 | $1.50 | $7.50 |
| Mistral Small 4 | $0.15 | $0.60 |
| Ministral 3 14B | $0.20 | $0.20 |
| Ministral 3 8B | $0.15 | $0.15 |
| Ministral 3 3B | $0.10 | $0.10 |
From our data, token economics vary sharply depending on model footprint: running 1 million input plus 1 million output tokens on Mistral Ministral 3 3B costs $0.20, while the same volume on Mistral mistral-medium-3-5 costs $9.00, which is 45 times as much. Mistral Large 4 sits between these at $0.70 per million input tokens and $2.10 per million output tokens.4
Operationalizing agents for security and trust
To satisfy governance requirements, Mistral Studio provides built-in hosting along with authentication and access controls, helping developers prepare applications for enterprise use faster. Agents are configured so they only reach the specific systems administrators explicitly connect them to.1
For ongoing reliability, Studio incorporates dedicated agent observability and evaluation tools. These monitoring capabilities give security teams visibility into model evaluations and execution paths, making it possible to measure performance benchmarks and audit automated decisions across mission-critical enterprise environments.1