Mistral offers open and portable generative AI models for developers and businesses. The models are designed to be usable and adaptable, with a variety of options for different use cases. The project is designed to promote trust, transparency and decentralization in AI development.
Mistral offers a suite of open-source models licensed under the Apache 2.0 license, including Mistral 7B, Mixtral 8x7B and Mixtral 8x22B. You can download them or use them on demand through the platform, which offers flexibility in how you deploy them. Optimized commercial models like Mistral Small and Mistral Large offer better performance and efficiency, and can be used on cloud partners and for self-deployment.
Among Mistral's features are:
Mistral models are geared for developers and businesses that want to automate large-scale text generation and processing, build internal assistants and develop custom AI applications. The technology is for AI-first startups and established companies, and the company's goal is to make AI more accessible to a broader range of people.
Mistral models are priced based on usage, with costs per million tokens. For example, the open-source Mistral 7B model costs $0.25 per million tokens, while the optimized Mistral Small model costs $1 per million tokens. Fine-tuning models also come with one-off training and storage fees.
Mistral prioritizes security, trust and moderation, protecting data privacy and encrypting data. The company is working on new developments to offer the most capable and efficient models, and the company says it offers a cost-effective way to integrate AI into businesses and industries.
Published on June 14, 2024
Analyzing Mistral...