Loading…
Model hub
Hugging Face
Open repository for downloading protein, DNA, RNA, single-cell and pathology foundation model weights, with managed inference endpoints alongside.
- Models served
- 262
- Offers
- Weights
- Coverage
- Protein 73Imaging 51Language model 45DNA & Gene 44+7 more
- Deployment
- Free download · managed endpoints
- Pricing
- Free + seat + hourly compute
Models you can download from Hugging Face
Most cited first5.4K2.2M4.2K
Protein structure prediction from a single sequence, with no multiple sequence alignment. Folds a 384-residue protein in 14.2 seconds on one GPU.
Protein
5.4K1.2M4.2K
Protein language model family from 8M to 15B parameters, used as a frozen sequence encoder whose representations encode atomic-level structure.
Protein
3.1K269.2K632
Sparse attention transformer that extends BERT to 8x longer sequences via random, local, and global attention, with genomic sequence applications.
DNA & Gene
2.5K—1.7K
Variational autoencoder pretrained on 74 million human single-cell transcriptomes from the CELLxGENE Census for batch correction and cell typing.
Single-cell
2K10.2K2.2K
Biomedical vision-language assistant for question answering on radiology and pathology images, adapted from LLaVA on PubMed Central captions.
PathologyLanguage model
How access works
Public repositories require no account: weights are cloned or downloaded directly, and credentials matter only for pushing changes or reading private repositories. Gated repositories are the exception: they require a signed-in account, consent to share username and email with the model's authors, and sometimes manual approval, with scripted downloads needing an access token. Authors can also block EU users by IP, so a public model is not necessarily an available one.
- Self-serve
- Free tier
- Public price list
- API access
- Open weights
- Self-hostable
- SSO / SAML
- SOC 2 Type II
- GDPR DPA
- Region selectable
What it costs
Downloading public weights is free, which for most use of the repository is the whole story. Costs begin with a seat or a machine. Seats are a flat monthly figure, rising per user across team and enterprise plans. Compute is hourly and published per accelerator, from a few cents for a small CPU endpoint to tens of dollars for a multi-GPU node. Storage bills per terabyte per month. Serverless inference carries a small monthly credit allowance, then passes through the provider's own rates.
Last verified
