France’s Mistral AI announced its first generative artificial intelligence model to compete with US leaders in AI, including Meta Platforms Inc. and Microsoft Corp.-backed OpenAI.
(Bloomberg) — France’s Mistral AI announced its first generative artificial intelligence model to compete with US leaders in AI, including Meta Platforms Inc. and Microsoft Corp.-backed OpenAI.
The Paris-based startup is releasing a “small-sized” language model with 7 billion parameters that will be made available for free to developers, it said in a statement on Wednesday.
“We have training methods that make us more efficient and two times less expensive to implement,” Mistral AI Chief Executive Officer Arthur Mensch said in an interview with Bloomberg. Large Language Models, or LLMs, are the base infrastructure for generative AI, a technology that creates content from videos to poetry out of simple user commands.
Four French companies are testing the model for applications including customer service, chatbots, summaries and marketing content generation, according to Mensch. The open-source model, called Mistral 7B, can work with English and code, and will be followed by larger models and commercial offers, he said.
Since the debut of OpenAI’s ChatGPT last year, investors have poured money into the potentially transformative technology. Microsoft has invested $13 billion into OpenAI.
Mistral, which has 18 employees, was founded this year by three researchers: Timothée Lacroix and Guillaume Lample — who both previously worked at Meta — and Mensch, formerly of Google’s DeepMind.
The company raised €105 million ($110 million) in an initial financing round announced in June that one of the largest seed rounds ever for a European generative AI company.
More stories like this are available on bloomberg.com
©2023 Bloomberg L.P.