Listen to this article
Mistral, a pioneering French AI startup, has recently unveiled its inaugural language model, which is not only free for all users but also boasts superior performance compared to other models of its size.
Open Models Gaining Traction
While many popular language models are accessible only via APIs, the trend towards open models is unmistakably on the rise. Mistral’s recent announcement of its first model, which is free from any usage restrictions, further solidifies this trend.
Mistral 7B: A Game-Changer in the AI World
The Mistral 7B model is now available for download through various channels, including a 13.4-gigabyte torrent. Additionally, Mistral has initiated a GitHub repository and a Discord channel to foster collaboration and address any issues users might encounter. The model’s release under the Apache 2.0 license ensures its unrestricted use, provided there’s proper attribution. This means that anyone, from individual hobbyists to large corporations, can utilize the model, given they have the necessary resources to run it.
Outperforming the Competition
Mistral 7B stands out as an enhanced version of other “small” large language models, such as Llama 2. It offers similar capabilities but at a significantly reduced computational cost. While foundational models like GPT-4 have broader capabilities, their high costs and complexities restrict their availability to APIs or remote access.
Mistral’s Vision for Open Generative AI
In a recent blog post, Mistral’s team expressed their ambition to lead the open generative AI community and elevate open models to state-of-the-art performance levels. The release of Mistral 7B, a culmination of three months of rigorous work, is a testament to their commitment. The team’s experience, having previously worked on similar projects at Meta and Google DeepMind, played a crucial role in this achievement.
The Fine Line Between “Free” and “Open Source”
While Mistral 7B is free for all users, it’s essential to differentiate between “free to use” and “open source.” The model was developed privately with private funding, and its datasets and weights remain confidential. Mistral’s business strategy revolves around offering the model for free, but those seeking deeper insights would need to opt for their paid products.
Mistral’s Future Endeavors
Mistral has hinted at their future plans, which include distributing their commercial offerings as white-box solutions, making both the code sources and weights accessible. The company is also actively working on hosted solutions tailored for enterprise deployment.
In conclusion, Mistral’s introduction of its free-to-use large language model marks a significant milestone in the AI industry. As open models continue to gain popularity, it will be interesting to see how companies like Mistral shape the future of AI.
Please email us your feedback and news tips at hello(at)dailycompanynews.com
