This latest funding comes just six months after Mistral AI’s initial $112 million seed round, which aimed to establish the company as a European competitor to OpenAI. Founded by former Google DeepMind and Meta employees, Mistral AI focuses on developing foundational models with an open-source approach.
Co-founder and CEO, Arthur Mensch, outlined the company’s vision: “Since our inception in May, we have pursued a clear trajectory: to become a European champion with a global vocation in generative artificial intelligence, based on an open, responsible, and decentralized approach to technology.“
Mistral AI introduced its first model, Mistral 7B, in September 2023. This large language model was designed to avoid direct competition with giants like GPT-4 or Claude 2. Notably, Mistral AI opted for a unique approach by making the model available as a free download instead of offering access through APIs. This allowed developers to run the model on their own devices and servers, empowering open experimentation and collaboration.
Meanwhile, Mistral AI actively engaged in shaping the EU’s AI Act, advocating for exemptions for foundational models. The company argued that regulation should primarily focus on use cases and companies directly developing products for end-users. The recent EU legislation ultimately reached a compromise, requiring transparency from companies working on foundational models. This necessitates the sharing of technical documentation and summaries of dataset contents.
Now, Mistral AI is shifting its focus towards monetization through its newly launched developer platform. This beta platform allows companies to pay for access to Mistral AI’s models via APIs. The platform initially offers two models:
- Mistral-tiny: This model corresponds to the original Mistral 7B release and is available for free download under the Apache 2.0 license.
- Mistral-small: This new model, known as Mixtral 8x7B, utilizes a router network to process input tokens efficiently, selecting the most appropriate parameters for generating responses. This technique allows for increased model complexity while maintaining cost and latency efficiency. Similar to Mistral-tiny, Mixtral-small is also freely available under the Apache 2.0 license.
Additionally, Mistral AI introduces a third model, Mistral-medium, which is exclusively available on the paid API platform. This model offers improved performance compared to Mistral AI’s other models and is strategically positioned to attract commercial users.
Overall, Mistral AI’s recent funding round and platform launch solidify its position as a major player in the AI landscape, particularly within the European market. The company’s dedication to open-source principles and responsible AI development aligns well with evolving regulations. By releasing models through diverse channels and launching its developer platform, Mistral AI demonstrates its commitment to shaping the future of artificial intelligence.
Add a Comment