IBL News | New York
French startup Mistral launched its latest open-source coding model, Devstral 2, yesterday.
This announcement follows the recent launch of the Mistral 3 family of open-weight models and Mistral Vibe, a new command-line interface (CLI) for code automation through natural language.
This announcement follows the recent launch of the Mistral 3 family of open-weight models and confirms Mistral’s
Devstral 2 requires at least four H100 GPUs or equivalent for deployment, as it weighs 123 billion parameters. However, the model is also available in a smaller version, Devstral Small, which, at 24 billion parameters, is deployable locally on consumer hardware.
Devstral 2, shipped under a modified MIT license, is currently free to use via the company’s API. After the free period, the API pricing will cost $0.40/$2.00 per million tokens (input/output) for Devstral 2.
Mistral has partnered with the agent tools Kilo Code and Cline to release Devstral 2 to users, while the Mistral Vibe CLI is available as an extension in Zed for use within the IDE.
Valued at $13.8 billion, this European unicorn [part of its team in the picture above] is aiming to close in on its larger American AI rivals and other coding-focused LLMs, such as Cursor and Supabase.