French startup Mistral AI has concluded its Series A funding, amassing €385 million (approximately $415 million USD) and achieving a valuation of around $2 billion. The company is also launching its commercial platform today, according to a report by Bloomberg.
Founded by ex-Google's DeepMind and Meta professionals, Mistral AI raised a substantial $112 million seed round just six months ago to establish itself as a European rival to OpenAI. It focuses on developing foundational models with an open technology approach, aiming to innovate within the AI domain.
The latest funding round is led by Andreessen Horowitz (a16z), with continued support from Lightspeed Venture Partners. Notably, a consortium of major investors, including Salesforce, BNP Paribas, CMA-CGM, General Catalyst, Elad Gil, and Conviction, has also participated in this remarkable funding initiative.
Arthur Mensch, Co-founder and CEO of Mistral AI, highlighted the company's vision, stating, "Since the creation of Mistral AI in May, we have been pursuing a clear trajectory, that of creating a European champion with a global vocation in generative artificial intelligence, based on an open, responsible and decentralized approach to technology."
Mistral AI recently unveiled its inaugural model, Mistral 7B, differentiated by its unique focus on a "small" dataset of approximately 7 billion parameters. The company diverged from conventional API accessibility by offering the Mistral 7B model as a free download, empowering developers to run it on their own servers and devices.
While the model operates under the Apache 2.0 license, enabling widespread utilization without usage constraints, its development process was conducted behind closed doors, utilizing proprietary datasets and undisclosed weights.
Moreover, Mistral AI actively participated in shaping discussions surrounding the EU's AI Act, advocating for a complete exemption for foundational models from stringent regulations. EU lawmakers recently reached a consensus, mandating transparency requirements for companies involved in foundational models, demanding the sharing of technical documentation and dataset summaries.
Looking ahead, Mistral AI plans to monetize its foundational models by opening up its developer platform in beta. This platform will enable companies to access Mistral AI's models via APIs, introducing the new Mixtral 8x7B model ("Mistral-small"), leveraging a router network technique to process input tokens efficiently.
With Mixtral 8x7B and Mistral-medium, the company aims to diversify its offerings, catering to different user needs while continuing to push the boundaries of AI innovation.