French generative artificial intelligence startup Mistral AI, led by a team of former Meta Platforms Inc. and Google DeepMind researchers, said today it has scooped up €105 million ($113 million) in seed funding to aid its push to become a “global player” in the market, just four weeks after it was founded.

Today’s round was led by Lightspeed Venture Partners and saw participation from a host of European capital firms, including JCDecaux Holding, Motier Ventures, La Famiglia, Headline, Exor Ventures, Sofina, First Minute Capital and LocalGlobe, plus individual investors such as former Google LLC Chief Executive Eric Schmidt, Xavier Niel and Rodolphe Saadé.

Based in Paris, Mistral is the latest startup looking to take on Microsoft Corp.- backed OpenAI LP and Google in building the base infrastructure for generative AI models, which can create new text and images from simple prompts. Today’s round is said to be one of the largest seed rounds ever raised by a European company, surpassing the $101 million nabbed by Stability AI Ltd. in October.

The size of the round is perhaps unsurprising given the pedigree of Mistral AI’s three co-founders. They include Timothée Lacroix and Guillaume Lample, who previously worked for Facebook parent company Meta Platforms Inc.’s AI research unit. There, Lample led the development of Meta’s most advanced large language model LLaMA.

The third co-founder is Arthur Mensch, who previously worked at Google’s AI research laboratory DeepMind. In an interview with TechCrunch, Mensch said he and the others began talking about the rapid advances of generative AI last year, at a time when OpenAI first began making headlines following the launch of ChatGPT, a chatbot that’s able to respond to questions in a humanlike manner

Mensch explained that they ultimately decided to get together and form Mistral AI, with a mission to bring generative AI to enterprises. “At the moment we have proof that AI is useful in some cases,” Mensch said. “But there are still too many workers in different fields being asked to be creative, and we need to figure this out for them. We want to give them tools that are easy to use to create their own products.”

Mistral AI hasn’t revealed exactly what it’s building, but it said its models and the datasets it uses to train them will all be open-sourced. The idea is that by using publicly available data, it can avoid legal issues, though its customers will also be able to contribute their own data.

Lightspeed’s partner Antoine Moyroud told TechCrunch one of the reasons he’s backing Mistral AI is its founders’ vision for generative AI, and knowledge of when and where it should be applied. He recalled that he and his partners met with a lot of AI startups, in India, Europe and the U.S. “What was lacking was, while they were strong technically, they hadn’t thought about the second order effects of LLM,” he said.

Mistral AI may have a strong sense of where it’s going, but it will be under pressure to deliver in what is fast becoming an extremely competitive landscape. Alongside OpenAI and Google’s Bard, Mistral AI faces competition from a host of well-funded startups, including Cohere Inc., Anthropic, AI21 Labs Ltd. and Stability AI.

Image: kenshinstock/Freepik