Late last month, Facebook parent Meta unveiled Llama 3.1, the world’s largest open-source model. With 405 billion parameters, it’s so big that even model libraries like Hugging Face need to scale up their systems to manage it and other models, as they continue to get bigger. That’s why Hugging Face, a $4.5 billion unicorn, has now acquired XetHub, a collaboration platform to help developers work with and build large-scale models, the company exclusively told Forbes. “Big models are here to stay,” Clem Delangue, CEO of Hugging Face, said in an interview. “What we want is to make the development of AI closer to what software engineering is — make it drastically faster.” The purchase price was undisclosed, but Delangue described it as the 8-year-old company’s largest acquisition. Previously, Hugging Face’s biggest buyout was its June acquisition of a dataset startup called Argilla for $10 million — a relatively humble price tag for the currently frothy world of AI. XetHub is a platform that allows people to collaborate when developing software using large models and datasets, helping them to see each other’s software changes and document their work. Particularly, it lets people iterate quickly when working with terabytes of data so they can proof their work without having to upload or download massive amounts of data for every potential change. The company, brd in Seattle, raised $7.5 million in its seed round, led by the venture firm Madrona. XetHub CEO Yucheng Low said the long term goal is to help Hugging Face be more efficient with ever-larger datasets, growing storage and more users. The XetHub platform, used by enterprise clients including the data analytics company Tableau, will be shut down and its features will be integrated into Hugging Face’s platform. When the integration is completed, Hugging Face thinks it will be able to host hundreds of millions of AI models and datasets in the next five years — far more than the 2.5 million it currently handles.

Full report : Hugging Face buys XetHub, a collaboration platform started by ex-Apple employees that raised $7.5 million to help developers work with and build large-scale models.