Mistral AI has introduced cloud-based remote agents in its Vibe coding environment, powered by Mistral Medium 3.5. The company says the model supports a 256,000-token context window and is priced at $1.50 per million input tokens and $7.50 per million output tokens.
Mistral AI has launched cloud-based remote agents in Vibe, its coding environment, powered by the company’s Mistral Medium 3.5 model.
In a company announcement titled “Remote agents in Vibe. Powered by Mistral Medium 3.5,” Mistral AI said the new Vibe remote agents are designed to run coding tasks in the cloud rather than only inside a local development session.
The announcement positions the feature as part of Mistral’s broader push into agentic coding workflows, where a model can help inspect code, make changes, and operate across larger software projects. According to Mistral AI, the remote agents are powered by Mistral Medium 3.5.
Mistral’s official model documentation describes Mistral Medium 3.5 as an “open v26.04 frontier-class multimodal model” intended for agentic and coding use cases. The same model card lists a 256,000-token context window, a capacity that is relevant for software development because larger context windows can allow models to consider more files, documentation, or project history in a single interaction.
Mistral AI states that API access to Mistral Medium 3.5 is priced at $1.50 per million input tokens and $7.50 per million output tokens. The same pricing is also listed in Mistral’s official model card for Mistral Medium 3.5.
7minAI News, which also covered the launch, reported that Mistral Medium 3.5 powers the Vibe remote agents and cited the same $1.50 and $7.50 per million-token API pricing. The outlet also described the model as a 128B dense model and reported a 77.6% score on SWE-Bench Verified. Those additional details were attributed in the report to the launch coverage, while Mistral’s own model documentation confirms the context window, use-case positioning, and pricing.
Remote coding agents are becoming a competitive area for AI model providers and developer-tool companies. Instead of simply answering prompts, these systems are intended to carry out software tasks over longer sessions, potentially including code edits, debugging, test execution, and repository-level reasoning.
Mistral’s move is notable because it connects a newly documented model, Medium 3.5, directly to a cloud-based coding product experience. The official documentation’s emphasis on agentic and coding use cases suggests that Mistral is targeting developers who need models that can handle long context, structured tool use, and software-engineering workflows.
The announced pricing also gives developers and organizations a clearer basis for comparing Mistral Medium 3.5 with other coding-oriented models. Token pricing does not by itself determine total cost, because usage depends on prompt size, output length, tool calls, and workflow design. But the published $1.50 input and $7.50 output rates provide a concrete benchmark for teams evaluating whether to run coding agents at scale.
Mistral AI has been positioning its model lineup around open and enterprise-oriented deployments, while also expanding into products that make its models directly usable for developers. Vibe remote agents appear to be part of that strategy: a product feature built around a model card that explicitly highlights multimodal, agentic, and coding capabilities.
For now, the main confirmed details are the launch of Vibe remote agents, their use of Mistral Medium 3.5, the model’s 256,000-token context window, and the API pricing published by Mistral AI. Further independent testing will be needed to assess how the remote agents perform across real-world repositories, languages, and software-engineering tasks.
Mistral AI has launched cloud based remote agents in Vibe, its coding environment, powered by the company’s Mistral Medium 3.5 model.
What Mistral announced In a company announcement titled “Remote agents in Vibe.
Powered by Mistral Medium 3.5,” Mistral AI said the new Vibe remote agents are designed to run coding tasks in the cloud rather than only inside a local development session.
Continue reading
