French company Mistral has launched Devstral 2. With this launch, the French AI startup is attempting to gain ground on larger competitors such as Anthropic and other specialized coding LLMs. Its not-so-secret weapon is an emphasis on following business context for optimal AI-generated code. In addition to the new Devstral model, there is also the playfully named Mistral Vibe, a command-line interface for code automation.
Mistral wants to catch up with the AI leaders, reports TechCrunch, which was the first to report on the new model. Anthropic’s Claude Code in particular is a popular model among programmers, but Devstral 2 aims to lure them to the French company. Mistral is also responding to the vibe coding trend with Mistral Vibe, a CLI tool for searching code in natural language, versioning, or executing commands. It integrates with Zed as an IDe extension.
Context awareness at the core
Mistral AI focuses on context awareness, which is particularly relevant for business applications. Like the AI assistant Le Chat, which remembers previous conversations and uses them for answers, Vibe CLI has persistent history. It also scans file structures and Git statuses to build context for behavioral information.
Devstral 2 does have substantial system requirements, however. It requires at least four (!) H100 GPUs or the equivalent and has 123 billion parameters. The more compact and less capable DevStral Small has 24 billion parameters. That is small enough to run on relatively conventional workstations without too many problems.
The licenses differ per model. Devstral 2 is covered by a modified MIT license, while Devstral Small uses Apache 2.0. In terms of API pricing, Mistral uses different rates: after the free period, Devstral 2 costs $0.40/$2.00 per million tokens (input/output) via the API, while Devstral Small charges $0.10/$0.30. Those who run it locally obviously do not have these ongoing costs, but as mentioned, the largest model requires fairly advanced equipment.
Partnerships for distribution
Mistral has entered into partnerships with agent tools Kilo Code and Cline for the release of Devstral 2. The launch marks a new step for the company, which has quickly become the European market leader. The startup is working on Mistral Compute, an AI cloud infrastructure built entirely in France.
The strategic partnership with and investment by ASML should help both the chip machine manufacturer and Mistral AI move forward. For ASML, it brings AI models to the production line, while Mistral gains access to capital and expertise.
Mistral AI’s valuation rose to €11.7 billion (approximately $13.8 billion) after a Series C financing round led by ASML. That €1.3 billion (approximately $1.5 billion) investment took place in September and made the Dutch chip giant the largest shareholder.