Mistral AI is making headlines once again after announcing a €600 million fundraising – a mix of equity and debt. This Series B was led by General Catalyst and DST. With this announcement, the young company reaches a milestone since it increases its valuation to 5.8 billion euros. Since June 2023 – where the start-up announced a seed round of 105 million euros – Mistral AI has come a long way and grown a lot.
Thus, last December, only six months after its creation, the star AI start-up in France announced to raise 385 million euros and saw its valuation skyrocket to two billion dollars (around 1.85 million euros). The prestigious Andreessen Horowitz fund and the leader in GPUs Nvidia entered its capital.
Expand its international presence, the United States as a priority
Today, the French AI benchmark founded by Arthur Mensch, Guillaume Lample and Timothée Lacroix, all three French, and former DeepMind and Meta employees, now has several dozen employees. “This new cycle puts us in a unique position to push the boundaries of AI and bring cutting-edge technology to everyone,” said co-founder and CEO Arthur Mensch.
The funds raised will be used to increase its computing capacity, recruit staff and expand its international presence, particularly in the United States, the company said in a statement. Reuters.
Its models distributed by the largest providers
Mistral AI can now boast of being the number one French reference in generative artificial intelligence and among the best in Europe. The first half of this year was thus dedicated to pursuing its strategy of multiplying platforms. The start-up was able to surround itself with leading companies, notably AWS, Databricks, Google, Microsoft, Salesforce, Snowflake, etc. Listening to one of its founders, Arthur Mensch, it is about“bringing generative artificial intelligence everywhere”. Asked about this during a technology event last March, he responded: “Not sticking to a single cloud provider is also our strategy.”
Thus, the open source models developed by the start-up are available on “all the big providers”, he specifies, as well as on the Hugging Face platform. “The idea is that our customers can choose the infrastructure and change over time.” Building on this success, Mistral AI has developed a certain number of models in recent months in order to develop its portfolio of solutions – and customers at the same time.
Several models launched in the space of a few months
One of its models has made a notable entry into the ecosystem, to say the least. Called Large and dedicated to high-level reasoning for complex tasks, it has, on several occasions, been considered a strong competitor to OpenAI's GPT-4 model.
More recently, it was its Codestral model, dedicated to code generation, which made headlines. One of the most popular use cases for LLMs is unsurprisingly code generation and the French start-up knows it. This open generative AI model is explicitly designed for code generation tasks and should help developers write and interact with code through a shared instruction and API. The model was trained on a diverse dataset of more than 80 programming languages, Mistral AI says, including the most popular ones, such as Python, Java, C, C++, JavaScript and Bash and is fluent in English.
Focus on personalization to please as many people as possible
At the same time, the French start-up launched services and an SDK to personalize its models. It is therefore possible to “fine-tune” its models from its platform. Three options are offered: self-service, managed or tailor-made, the objective always being the same: pushing for ultra-personalization to reach a greater number of companies and, more broadly, sectors just as varied.
Selected for you