Mistral AI Unveils Mistral Medium 3.5 and Cloud-Based Coding Agents

Mistral AI has introduced its new flagship model, Mistral Medium 3.5, and launched cloud-based coding agents in Vibe, according to a company announcement. The model is now the default for both the Mistral Vibe environment and the Le Chat assistant.

Mistral Medium 3.5 is a dense 128 billion parameter model with a 256,000 token context window. It merges instruction following, reasoning, and coding into one unified system. The model supports configurable reasoning effort per request and includes a vision encoder trained to handle variable image sizes. It is released under a modified MIT license with open weights available on Hugging Face.

The new Vibe remote agents allow coding sessions to run asynchronously in the cloud. Developers can start tasks from the Vibe command-line interface or directly in Le Chat. These sessions continue independently and can be monitored or teleported between local and cloud environments. The agents integrate with GitHub, Linear, Jira, and Slack to manage code, tasks, and reports.

Mistral also added a Work mode in Le Chat, currently in preview. This mode enables the assistant to handle complex, multi-step workflows such as research, analysis, and cross-tool coordination. Actions and tool calls are visible to users, with explicit approvals required for sensitive operations. The new features are available on Pro, Team, and Enterprise plans, with API pricing set at $1.5 per million input tokens and $7.5 per million output tokens.