Mistral AI unveiled Devstral 2, a 123‑billion‑parameter open‑source coding model for developers, enterprises, and individual users, alongside Devstral Small 2, a 24‑billion‑parameter variant geared toward broad accessibility.

Mistral also introduced Vibe CLI, a natural‑language driven tool that can perform project‑wide actions like refactoring and file management, usable as a terminal utility or through IDEs.

The pricing for API usage is $0.40 per million input tokens and $2.00 per million output tokens, with claims of cost advantages over Claude Opus 4.5 in certain scenarios.

TechCrunch reported on the release, with a full report link available for deeper details.

Mistral is expanding through partnerships with Kilo Code and Cline, and Vibe CLI integrates with IDEs via a Zed extension; community engagement and hiring channels are highlighted.

Strategically, Mistral positions open‑weight, privacy‑focused AI as a European counterweight to walled‑garden ecosystems, amid moves by Google Cloud with Replit and Anthropic’s investments.

Industry implications include faster AI adoption across software development, evolution of agent‑based workflows, and potential shifts in developer tools markets, alongside considerations around hardware, security, ethics, and licensing.

The platform emphasizes “vibe coding” over traditional chat assistants, enabling multi‑file orchestration with persistent project context.

The release could intensify the AI agent wars in developer tooling, driving competitors to innovate UIs and integrations and spur open‑source ecosystem growth around Mistral technologies.

Sponsorship and related tech‑news coverage accompany the article, indicating a broader sponsorship format rather than a single feature piece.

Early adopters report strong usage and high tool‑calling success for Devstral 2, with benchmarks showing competitive or superior performance to larger proprietary rivals in several tasks.

This release follows Codestral and Devstral precedents and sets the stage for Mistral 3, an ecosystem‑oriented, edge‑friendly line of open‑weight models focused on transparency and developer experience.