AI Briefing
KO

Mistral Vibe, Always Connected to Your Terminal

·2026.01.27 09:00

Key point

Mistral unveiled the terminal-based coding agent Vibe 2.0, along with Devstral 2 and a paid API policy.

Details

With the release of Mistral Vibe 2.0, the agent that automates coding work in the terminal has evolved another step further. This upgrade is built on the Devstral 2 model family, and its core focus is letting teams and individuals configure the agent to fit their own way of working.

The main new features are as follows.

  • Custom subagents: You can create subagents specialized for specific tasks—like deployment scripts, PR reviews, or test generation—and call them whenever needed.
  • Multi-choice clarifications: When intent is unclear, instead of proceeding arbitrarily, the agent presents multiple options to confirm first.
  • Slash-command skills: Use / commands to call up predefined workflows such as deployment, linting, or documentation generation.
  • Unified agent modes: Create custom modes that bundle tools, permissions, and behavior, so you don't need to swap tools every time the context changes.
  • Continuous updates: Bug fixes and improvements to the Vibe CLI are continuously and automatically rolled out.

This service is available on Le Chat Pro and Team plans, and once usage limits are exceeded, billing continues on a pay-as-you-go basis at API rates. Devstral 2 is now transitioning to paid API access, though free experimental use is available on the Experiment plan of Mistral Studio.

As for pricing, Devstral 2 costs $0.40/M tokens for input and $2.00/M tokens for output, while Devstral 2 Small costs $0.10/M tokens for input and $0.30/M tokens for output. For enterprise customers, Mistral also offers fine-tuning for internal languages and DSLs, reinforcement learning using their own environments, and code modernization services to migrate legacy codebases to modern stacks.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.