Mistral AI: Who's Behind Europe's Main Open-Model Rival to OpenAI

Mistral AI is a French company founded in April 2023. It has raised hundreds of millions of dollars and become OpenAI's main European competitor in the large language model market.

Who created Mistral AI

Three researchers from leading AI labs founded the company: Arthur Mensch (DeepMind), and Guillaume Lample and Timothée Lacroix (both from Meta AI Research). All three had worked on world-class language models, giving the startup a rare concentration of expertise from day one.

The company is headquartered in Paris, which immediately positioned Mistral as Europe's answer to the American AI giants. Key funding milestones:

• April 2023 — company founded
• June 2023 — $113 million seed round at a $260 million valuation
• December 2023 — $415 million Series A at a $2 billion valuation
• June 2024 — $640 million Series B at a $6 billion valuation

What models has Mistral released

Openness is the company's main competitive pitch. Its first public model, Mistral 7B, shipped in September 2023 under the Apache 2.0 license and quickly caught developers' attention: with seven billion parameters, it delivered results comparable to larger closed models. It could even run on a consumer GPU.

In December 2023, the company released Mixtral 8x7B, the first widely known openly available implementation of the Mixture of Experts architecture. The model activates only a subset of its parameters for each request, which cuts compute costs without a significant loss in quality.

In parallel, Mistral is building a closed commercial lineup: Mistral Small, Medium, and Large, offered through its La Plateforme API platform. For end users, there's Le Chat, the company's own chat interface and a direct ChatGPT competitor. Enterprise customers can deploy the models on-premises on their own infrastructure.

Why Mistral matters for the market

With American (OpenAI, Anthropic, Google) and Chinese (DeepSeek) labs dominating the field, Mistral remains the main proof that Europe can compete at the frontier-model level. The French government publicly backs the company, and the European Commission sees it as a strategic asset for digital sovereignty.

"Our goal is frontier AI in everyone's hands" —

Mistral has used this principle since day one to justify its choice of open weights.

Open models let organizations run AI locally without sending data to external clouds. For European enterprise and government customers operating under strict GDPR and sector-specific regulatory requirements, this opens doors to segments where OpenAI struggles to get in. Open weights also make it easier to audit models for compliance with the EU AI Act.

What this means

Mistral AI shows that you can compete with the American AI giants through openness and efficiency, not just through scale and closed ecosystems. The company remains one of the key players shaping the future balance between open and proprietary AI.

FAQ

Are Mistral's models free?

The base models, Mistral 7B and Mixtral 8x7B, are distributed under the Apache 2.0 license: you can download and use them for free, including in commercial projects. More powerful models (Mistral Large) are available only through the paid API.

How is Mistral different from OpenAI?

Mistral bets on open-source base models and European jurisdiction, while OpenAI keeps its flagship models closed. That said, Mistral also has a closed commercial lineup, so the two companies compete head-to-head in the API access segment.

*Meta is designated an extremist organization and banned in Russia.