Hugging Face
Hugging Face is a platform and community hub hosting over a million open source machine learning models, datasets and demo apps (Spaces), alongside open source libraries like Transformers. It offers free public hosting plus paid PRO, Team and Enterprise plans, and usage-based pricing for private storage, Inference Endpoints and Spaces hardware. Nvidia confirmed on 3 September 2026 it is acquiring Hugging Face for about $12.9 billion; Hugging Face continues to operate under its own brand while the deal has not yet closed.
Facts checked on against the sources listed at the end of this page.
- Pricing
- FreemiumPaid from $9 a month for a PRO account
- Open source
- YesOpen source libraries (e.g. Transformers) are Apache-2.0; hosted models and datasets carry their own individual licenses
- GitHub stars
- 168,260+1,524 in the last 30 days
- npm downloads a week
- 440kup 2% on the week before
- Latest release
- v5.211 day ago
- Releases in 90 days
- 10about 3 a month
A good fit for
- Finding, downloading and sharing open source machine learning models and datasets
- Hosting a demo app for a model with Spaces, without managing your own server
- Teams that want managed, billed-by-the-hour Inference Endpoints for a hosted model
- Organizations that need access control and compliance features for private model and dataset storage
Not a good fit for
- Teams that want long-term certainty about ownership and roadmap, given the pending Nvidia acquisition
- Very high-volume production inference where a dedicated, self-managed GPU cluster may be cheaper than hosted Inference Endpoints
- Anyone who needs guaranteed support or licensing terms for every individual model on the hub, since licenses vary by model
Hugging Face pricing
The hub is free for public use. PRO accounts are $9 a month, Team is $20 a month per user and Enterprise Hub is $50 a month per user. Private storage, Inference Endpoints and Spaces hardware are billed separately by usage.
| Plan | Price | What you get |
|---|---|---|
| Free | $0 | Unlimited public models, datasets and Spaces, free CPU and limited ZeroGPU quota. |
| PRO | $9 a month | More private storage, more inference credits, higher ZeroGPU quota, ability to host private Spaces. |
| Team | $20 a month per user | SSO, data location control, audit logs, granular access control, repository analytics. |
| Enterprise Hub | $50 a month per user | Everything in Team plus SCIM provisioning, advanced security, managed billing and dedicated support. |
Prices in USD, checked on 27 September 2026. Current prices on huggingface.co.
Main features
- Model and dataset hub
- Public hosting for over a million models and datasets, free to download and publish to.
- Spaces
- Host demo apps built with Gradio, Streamlit or Docker, with free CPU and ZeroGPU tiers or paid dedicated hardware.
- Transformers and Datasets libraries
- Open source Python libraries, licensed Apache-2.0, for using and training models from the hub.
- Inference Endpoints
- Deploy a model as a dedicated, billed-by-the-hour API endpoint, from CPU instances up to multi-GPU tiers.
- Inference Providers
- Route inference calls to third-party providers hosting popular open models, billed through Hugging Face.
- Enterprise Hub
- SSO, audit logs, SCIM provisioning and managed billing for organizations, on top of Team-plan access controls.
About Hugging Face
Hugging Face hosts a public hub of machine learning models, datasets and Spaces, small demo apps that anyone can run and fork, alongside widely used open source libraries such as Transformers and Datasets. Most of the hub is free to browse, download and publish to.
Paid plans add private storage, more compute and team features: a PRO account for individuals, a Team plan with SSO and access controls, and an Enterprise Hub plan with SCIM, advanced security and managed billing. Inference Endpoints and Spaces hardware are billed by the hour based on the CPU or GPU tier chosen, and storage beyond the free allowance is billed per terabyte with volume discounts.
Nvidia confirmed on 3 September 2026 it is acquiring Hugging Face for roughly $12.9 billion. As of this check the acquisition has not closed and Hugging Face continues to run independently under its own brand.
- Made by
- Hugging Face (pending acquisition by Nvidia)
- Free plan
- Yes: unlimited public model, dataset and Space hosting, free CPU and limited ZeroGPU compute
- Available on
- Web hub, Python (Transformers, Datasets), REST API, Inference Endpoints, Spaces (Docker/Gradio)
Usage and activity
Weekly npm downloads of @huggingface/inference, last 12 months.
Recent releases
Hugging Face alternatives
Hugging Face compared
Questions about Hugging Face
Is Hugging Face being acquired?
Nvidia confirmed on 3 September 2026 that it is acquiring Hugging Face for about $12.9 billion. The deal had not closed yet as of this check.
Is Hugging Face free to use?
Yes, browsing, downloading and publishing public models, datasets and Spaces is free. Paid plans add private storage, more compute and team features starting at $9 a month.
What is a Space on Hugging Face?
A Space is a hosted demo app, built with Gradio, Streamlit or Docker, that runs on free or paid compute and can be shared or forked by others.
Are Hugging Face models open source?
Hugging Face's own libraries like Transformers are open source under Apache-2.0. Models and datasets hosted on the hub each carry their own license set by whoever published them.
Sources and links
- Hugging Face pricing
- Nvidia confirms it will buy Hugging Face for $12.9 billion, TechCrunch
- Hugging Face, Wikipedia
Work on Hugging Face? Claim this listing. Something wrong or out of date? Suggest a correction. Stats come from the public GitHub and npm APIs and update every day.