The open home of machine learning: 3 million models, 500,000 datasets and 1 million apps, with hosted inference, Spaces and the Transformers libraries.
Hugging Face is the platform where the open machine learning community shares models, datasets and applications. More than 18 million developers, researchers and creators use it to publish and discover open source and open-weight models, host demos as Spaces, and run models through inference endpoints and third-party providers. Its libraries, including Transformers, Diffusers and Datasets, are the standard way most open models are loaded and fine-tuned. Public work is free; paid plans add private repositories, compute and team features. In September 2026 NVIDIA agreed to acquire Hugging Face, saying the platform will stay open and that NVIDIA compute will not be required to build on or deploy through it.
Hugging Face is the platform where the open machine learning community shares models, datasets and applications. More than 18 million developers, researchers and creators use it to publish and discover open source and open-weight models, host demos as Spaces, and run models through inference endpoints and third-party providers. Its libraries, including Transformers, Diffusers and Datasets, are the standard way most open models are loaded and fine-tuned. Public work is free; paid plans add private repositories, compute and team features. In September 2026 NVIDIA agreed to acquire Hugging Face, saying the platform will stay open and that NVIDIA compute will not be required to build on or deploy through it.
Tags: AI, Open Source, Developer Tools, LLM, Machine Learning
Try Hugging FacePublic models, datasets and Spaces are free to host and use. Paid plans add private repositories, more compute for Spaces and inference, and organisation features.
Spaces are hosted apps, usually Gradio or Streamlit, that let you try a model in the browser or share a demo of your own.
NVIDIA agreed to acquire Hugging Face in September 2026 for about $12.93 billion. NVIDIA says the platform will remain open: developers choose their models, frameworks, clouds and hardware, and NVIDIA compute is not required.