{"architecture":{"family":"GPT-style transformer with gene and expression-value tokens","input":"single-cell expression matrix (h5ad)","output":"cell/gene embeddings; predicted expression","pretraining":"generative masked prediction of expression values"},"article":null,"cancer_slugs":["pan-cancer"],"category":"genomics","confidence":"high","datasets":[],"developer":"Bo Wang Lab, University of Toronto","evaluation":[],"hf":null,"kind":"foundation","license":"MIT","limitations":"- Perturbation predictions are validated mainly on Perturb-seq lines, not patient tumours.\n- GPU memory scales with gene vocabulary; subsample genes for large atlases.","links":{"demo":null,"docs":null,"doi":"10.1038/s41592-024-02201-0","github":"https://github.com/bowang-lab/scGPT","huggingface":null,"paper":"https://www.nature.com/articles/s41592-024-02201-0","pmid":null},"modalities":["single-cell","transcriptomics"],"name":"scGPT","notable_uses":"","openness":"open-weights","regulatory":{"intended_use_en":"Research tool.","intended_use_pl":"Narz\u0119dzie badawcze.","source_url":"https://github.com/bowang-lab/scGPT","status":"not-applicable"},"regulatory_status":"not-applicable","release_date":"2024-02-08","run_snippet":"","settings":["basic-research"],"slug":"scgpt","sources":[{"label":"Cui H et al. scGPT: toward building a foundation model for single-cell multi-omics using generative AI. Nat Methods 2024","url":"https://www.nature.com/articles/s41592-024-02201-0"},{"label":"GitHub \u2014 bowang-lab/scGPT","url":"https://github.com/bowang-lab/scGPT"}],"summary":"Generative pretrained transformer for single-cell multi-omics, trained on over 33 million cells, supporting cell-type annotation, batch integration, perturbation response prediction and gene-network inference.","tasks":["single-cell","gene-expression","feature-extraction"],"training":{"size":"33M cells","summary_en":"33 million cells from CELLxGENE and other public collections; organ- and tissue-specific checkpoints (including a pan-cancer checkpoint) are released.","summary_pl":"33 mln kom\u00f3rek z CELLxGENE i innych zbior\u00f3w publicznych; wydano punkty kontrolne dla narz\u0105d\u00f3w i tkanek (w tym pan-nowotworowy)."},"updated_at":"2026-09-05T22:26:01.064388","url":"/ai-oncology/models/scgpt","usage":{"library":"PyTorch (scgpt package)","notes_en":"Checkpoints (whole-human, pan-cancer, per-organ) are linked from the repo README; expects scanpy AnnData input."},"verified_at":"2026-09-05T22:26:01.064048","verified_by":"editorial","version":null,"what_it_does":""}