Mistral AI unveiled a groundbreaking AI model on June 10, 2025, touted as “Europe’s first AI reasoning model.” The French startup said the new system employs chain-of-thought reasoning techniques – generating intermediate logical steps – to solve complex problems.
This approach marks a departure from simply scaling up model size, which is reaching practical limits, and could open a new path in advanced AI capabilities.
Mistral’s latest release positions the company to keep pace with American and Chinese rivals: OpenAI pioneered such reasoning models the previous year, and China’s DeepSeek open-sourced a reasoning AI in January. “For Mistral, valued at $6.2 billion, an industry shift away from pure scaling gives a window to catch up against better-capitalized rivals,” Reuters noted.
In a bid for differentiation, Mistral is embracing openness and multilingual capability. The company launched two versions – an open-source “Magistral Small” model (24 billion parameters) and a more powerful proprietary “Magistral Medium” for enterprise use.
Unlike many competitors’ closed systems, Mistral’s smaller model is available for download on Hugging Face and can reason in English, French, Spanish, Arabic, and Chinese.
This transparency echoes Mistral’s strategy of leveraging its European roots: it has open-sourced earlier models (like Mistral 7B) to showcase technological talent while courting support from officials including French President Emmanuel Macron.
Experts say reasoning models could be pivotal for AI’s next leap, and Mistral’s move signals Europe’s intent to innovate rather than merely follow Silicon Valley. By combining home-grown research with an open ethos, the launch bolsters Mistral’s reputation as Europe’s top contender in foundational AI models.
