Mistral AI launched the Mistral Studio platform for deploying AI agents alongside the release of the Mistral Large 4 model
Alongside the release of the Mistral Large 4 model weights, Mistral AI launched the Mistral Studio platform for deploying agents, workflows, and applications in the cloud, VPC, and on-premises. Studio is already used in production by ABANCA, BNP Paribas, Accenture, and other companies.
Mistral AI has expanded its enterprise platform Mistral Studio with managed agents, applications, and new features for workflows, enterprise context, agent observability, and evaluations. According to the company, Studio enables deployment in the cloud, in a private VPC, or on-premises without dependence on a single provider. The new Mistral MCP server lets developers build an agent using the Mistral Vibe tool or their own coding agent, test it locally, and deploy it to the Studio environment with a single command. According to the company, the platform is already used in production by ABANCA, BNP Paribas, Accenture, Alten, Capgemini, CMA CGM, and Tata Consultancy Services.
Studio was released alongside the weights of the improved Mistral Large 4 model (nicknamed Le Chonk), which Mistral AI published following the public preview. It is a natively multimodal model with a trillion parameters, focused on coding, cybersecurity, and business reasoning. According to the company, the model achieved 41 percent on the Terminal-Bench 4.0 benchmark, 13 percentage points more than the preview version, and 48 percent on the Artificial Analysis Coding Agent Index.
The company describes specific real-world deployments of Studio: according to the company, an agent for investigating production incidents saves engineers hundreds of hours per month, and a workflow for triaging ideas and customer feedback saves the product department approximately 35 hours per week. These are claims made by the company, not independently verified figures.
Why it matters
Mistral Studio moves Mistral AI from offering a standalone model to a full enterprise platform that competes with agentic automation tools from other providers. For companies, this means the ability to deploy AI agents and workflows even in a closed environment (VPC, on-premises) without dependence on a single cloud provider, which is relevant for regulated industries or organizations dealing with data sovereignty.
Release card
Mistral Large 4
Mistral AI
- Specifications
- 1 trillion parameters
- Inputs
- natively multimodal
- Availability
- The model weights have been released for download, and the model can be deployed independently (self-hosted), or through the Mistral Studio platform in the cloud, a private VPC, or on-premises.
- FrontierCode 2.0 19,7 % Cognition's benchmark assesses whether a repository maintainer would actually merge a change created by the agent.
- Terminal-Bench 4.0 41 % This represents a 13 percentage point improvement over the public preview version of the model on terminal tasks, putting it, according to the source, in a leading position among open-weight models.
- Artificial Analysis Coding Agent Index 48 % According to the source, the result is comparable to the GLM-5.3 model among open-weight models, but with 47 percent fewer tokens per task.
- Artificial Analysis Cyber Index 51 % The index measures the model's ability to find, reproduce, and fix vulnerabilities in code without breaking existing functionality.
- Průměrná míra odmítnutí u škodlivých kybernetických požadavků 96,6 % According to the source, this is the highest measured refusal rate for malicious requests among the open-weight models the company measures.
- agentic and terminal coding tasks with lower token consumption
- defensive cybersecurity, i.e., finding, reproducing, and fixing vulnerabilities
- enterprise tasks involving creating and editing spreadsheets and financial workflows
- the source cites the risk of capabilities being potentially misused for real-world attacks, which is why the company continues to adjust its guardrails
- the specialized variant for penetration testing and red-teaming is available only to vetted, selected partners, not publicly
The source compares the model with other open-weight models, specifically GLM-5.3, Kimi K3, and DeepSeek V4.1 Flash, in coding and in Surge Code's blind pairwise evaluation, where, according to the source, the model ranked behind Claude Opus 5.5.
The card summarizes information from the article and any dated corrections, with a link to the original source. It is not our assessment of the model. It does not yet have a dedicated editorial profile. Model selection and other announcements →
What was added since the original report
Verified updates
-
Mistral Studio is a new platform released alongside Mistral Large 4, focused on enterprise AI; Studio includes managed agents, workflows, enterprise context, and agent observability; Studio supports deployment on the cloud, VPC, and on-premises without dependence on a single provider; the Mistral MCP server enables integration with your own systems; ABANCA, BNP Paribas, Accenture, and 7 other companies already use Studio in production
- Mistral Studio is a new platform released alongside Mistral Large 4, focused on enterprise AI
- Studio includes managed agents, workflows, enterprise context, and agent observability
- Studio supports deployment on the cloud, VPC, and on-premises without dependence on a single provider
- The Mistral MCP server enables integration with your own systems
- ABANCA, BNP Paribas, Accenture, and 7 other companies already use Studio in production
Two audiences, two different impacts
What this means
For individuals
Thanks to the Mistral MCP server and the Mistral Vibe tool, developers can build agents and workflows locally and deploy them with a single command, which shortens the path from prototype to production deployment.
For a business
Companies gain a platform for deploying agents, workflows, and applications across the cloud, their own VPC, and on-premises, which, according to the client examples cited, reduces the time required for routine processes such as ticket triage or incident investigation.
ProcessesCheck the original
Event sources
clearly official source · 1 publisher, 0 independent. We count feeds from the same owner only once.