Community GPU Grants

Community GPU Grants

Provides AI builders with in-kind compute to run demanding model training and demonstration projects.

Hugging Face Community GPU Grants are an in-kind compute programme operated by Hugging Face, the open-source AI company and model-hosting platform that hosts over one million public models, datasets, and Spaces on huggingface.co. The programme awards free GPU hardware upgrades directly to qualifying Spaces, enabling researchers, developers, and organizations to run compute-intensive machine learning demos without paying hourly cloud GPU rates. There are no application deadlines; applications are accepted on a continuous rolling basis and no cash is transferred at any point. The default award hardware is ZeroGPU, powered by Nvidia RTX Pro 6000 Blackwell with up to 96 GB of GPU memory, which is the standard tier assigned to most approved grants. In some cases an A10G instance is granted instead. A100 hardware is explicitly not available under standard community grants. At standard retail pricing, the equivalent continuous-use compute value ranges from approximately USD 720 to USD 1,800 per month, though no monetary value is formally assigned or paid. Grants are temporary: Hugging Face may revoke GPU access if Space usage drops to very low levels, and no guaranteed duration or formal renewal process is published. The programme is open to individuals, academic teams, nonprofits, and companies worldwide without geographic restriction, provided they have an active public Space on huggingface.co. The application is embedded directly inside the Space's hardware repository settings under the sleep time settings section; there is no external portal or separate submission form. Hugging Face staff review applications on a rolling basis, with observed approval timelines of approximately two days in documented examples, though no formal service-level agreement is stated. Strong preference is given to academic research projects, open-source community demos, and projects with demonstrated traction such as high visitor counts, GitHub stars, or associated research papers.