Hugging Face is an American company, headquartered in New York City with a large Paris engineering presence, that builds open infrastructure for Machine learning. It began in 2016 as a teenager focused chatbot app, but pivoted in 2018-2019 toward open source tools after its Natural language processing library, Transformers, gained traction among researchers working with the Transformer (architecture) architecture. The company is often described informally as "the GitHub of machine learning" because its Hub hosts hundreds of thousands of Foundation model checkpoints, datasets, and interactive demo apps called Spaces, most of them uploaded by outside developers and labs rather than by Hugging Face itself.
The company was co-founded by Clement Delangue, Julien Chaumond, and Thomas Wolf. Delangue has served as chief executive since founding and has become a prominent public advocate for open Open-weights models releases as a counterweight to closed labs such as OpenAI and Anthropic.
Products and platform
Hugging Face's core open source libraries include Transformers (a unified interface for loading and running pretrained models across PyTorch and TensorFlow), Datasets, Tokenizers, and Diffusers for Diffusion model pipelines. The Hub layers a hosting and versioning service on top, similar in spirit to a code repository, along with model cards documenting training data, intended use, and limitations. Paid offerings include Inference Endpoints for deploying models as APIs, AutoTrain for no-code Fine-tuning, and enterprise support contracts, which fund the mostly free public platform.
Major labs including Meta AI, Mistral AI, Stability AI, DeepSeek, and the makers of Qwen routinely publish weights on the Hub, and it became a standard distribution channel after the 2023 leak of Llama weights spread rapidly through the platform. Hugging Face also hosts leaderboards that let the community compare open models on standard AI benchmark suites.
Funding and business model
Hugging Face has raised capital from investors including Google, Amazon, Nvidia, Salesforce, and Sequoia Capital, reaching a valuation reported at around 4.5 billion dollars in a 2023 funding round. Its business model depends on converting a portion of its large free user base, drawn by open access to models such as BERT descendants and community fine-tunes, into paying enterprise customers for compute and support, a pattern common among open source infrastructure companies.
Role in the AI ecosystem
Because so much open research and so many derivative models pass through its Hub, Hugging Face functions as a de facto standards body for how models are packaged, documented, and licensed in the open source AI community. It has published research on topics including dataset transparency and model evaluation, and its Model Card format has been widely adopted as a norm for documenting a model's training data and known limitations, an important input for debates in AI ethics and AI governance over transparency requirements for released models.