French AI startup Mistral is offering new customization options for its AI models, allowing developers and enterprises to fine-tune the models for specific use cases. This move is aimed at providing more flexibility and tailored solutions for users.
One of the options is a self-service approach, which includes the release of a software development kit (SDK) called Mistral-Finetune. This SDK allows users to fine-tune Mistral’s models on various platforms, from workstations to small datacenter nodes. The SDK is optimized for multi-GPU setups but can also be scaled down to a single GPU for smaller models. For example, fine-tuning on a dataset like UltraChat takes around half an hour using Mistral-Finetune across eight GPUs.
For those who prefer a more managed solution, Mistral also offers fine-tuning services through its API. These services are compatible with two of Mistral’s models, Mistral Small and Mistral 7B, with plans to add support for more models in the near future. This option provides developers and companies with a convenient way to customize the models without the need for extensive technical knowledge.
In addition to the SDK and API options, Mistral is introducing custom training services exclusively for select customers. This service allows organizations to fine-tune any Mistral model using their own data, enabling the creation of highly specialized and optimized models for specific domains. This approach provides even greater customization and control over the AI models, catering to the unique needs of each organization.
Mistral’s decision to offer these customization options comes as no surprise considering the competitive landscape of the generative AI space. The company, which is seeking significant investment, aims to increase its revenue and solidify its position in the market. While Mistral has released multiple generative models and paid APIs since its first model launch in September 2023, specifics regarding user numbers and revenues have not been disclosed.
In conclusion, Mistral’s introduction of AI model customization options, including self-service SDK, managed fine-tuning services, and custom training services, demonstrates the company’s commitment to providing tailored solutions for developers and enterprises. These options offer flexibility, convenience, and specialized customization to meet the diverse needs of users. As Mistral continues to face competition in the generative AI space, these customization options can help differentiate the company and drive revenue growth.