Open-weight histology foundation models on one bench.
Pick a clinical or preclinical model, give it tiles or a whole slide, and get the output its training was built for: embeddings and cluster maps, zero-shot class maps, rodent toxicologic pathology QA, virtual multiplex marker maps, multi-organ segmentation, or predicted spatial expression — with direct links to every paper. Models run on your own GPU server; this page never sees your images.
Run a model
Jobs queue on your server and run one at a time on its GPU.
Results
Jobs
Papers, citations & benchmarks
Direct DOI / arXiv / bioRxiv links and one-click BibTeX entries for every foundation model in the registry, alongside key clinical and rodent preclinical benchmarks and closed-weight reference models.
Open-access oncology data — organ-specific & pan-cancer
Cancer-only open deposits across Zenodo, Nature Scientific Data, Hugging Face Datasets, NCI IDC/TCIA and Weizmann 3CA / CELLxGENE — split into Pan-Cancer and Organ-Specific cohorts with priority coverage of Brain, Lung, Breast and Prostate cancers across matched and unmatched modalities.
Connect a backend
GitHub Pages only serves this page. The models run in the small Python server in backend/, on whichever machine has the GPU.
Server status
Connect to see GPU, packages and which models are ready.