Hugging Face recently acquired XetHub. With this, the hosting platform for LLMs is buying a start-up that helps developers manage the files they need to run AI projects.
The exact amount of the acquisition was not disclosed, but it is said to be Hugging Face’s largest to date. Among other things, this means that the acquisition amount is higher than the $10 million the LLM hosting platform paid for AI training specialist Argilla in June this year.
Code file management platform
With XetHub, Hugging Face gets its hands on a start-up that provides a platform for developers to store the code files and other technical solutions they use for AI projects. The main focus here is machine learning.
In addition, this platform offers several other tools to manage the ML stack. These include version control, keeping an eye on assets, making changes in a single place and more. The platform also offers (collaboration) productivity tools that should make it easier for developers to work with these specific files, edit them, improve performance and make better comparisons with other tooling.
Integration into the proprietary platform
Hugging Face wants to integrate XetHub’s technology into its own platform, primarily to improve the storage system of its hosting platform for LLMs. This should, among other things, improve the storage of LLMs in the open-source tool Git and remove existing limitations.
More specifically, XetHub helps, among other things, update LLMs faster. It does this by breaking up LLMs and underlying datasets into smaller pieces. When developers want to release an update, they only need to update those specific pieces instead of the entire file, speeding up update times. Another part of XetHub’s technology helps visualize the architecture of neural networks to make them more understandable to developers.
Finally, the acquisition is interesting for the AI hosting platform to start earning more from its commercial variant, Enterprise Hub. XetHub’s technology can add nice value for this purpose.
Also read: Hugging Face pushes for development of locally running AI with SmolLM