Model catalog
Browse models below. Running inference requires a free account.
A Mistral-7B model further pre-trained on PubMed Central open-access text for biomedical question answering. Use it like a chat assistant for biomedical topics.
A Segment Anything model fine-tuned on medical images across many modalities. Upload an image and draw a bounding box; MedSAM returns a segmentation mask overlay.
A DenseNet-121 trained across multiple public chest X-ray datasets. Upload a frontal chest X-ray to get probability scores across common pathologies.
Microsoft's GPT-2-style model pre-trained on biomedical literature. Enter a prompt and it continues it with biomedical text.
A Llama-3-8B model fine-tuned for biomedical QA. Chat with it about biomedical topics.
PubMedBERT is a BERT encoder for biomedical/clinical text. Enter a sentence with a [MASK] token and it predicts the most likely fillers.
SciBERT is a BERT encoder for biomedical/clinical text. Enter a sentence with a [MASK] token and it predicts the most likely fillers.
Bio_ClinicalBERT is a BERT encoder for biomedical/clinical text. Enter a sentence with a [MASK] token and it predicts the most likely fillers.
BioLinkBERT is a BERT encoder for biomedical/clinical text. Enter a sentence with a [MASK] token and it predicts the most likely fillers.
Clinical-text encoder (NER, relation extraction backbone). Returns an embedding of your text.
Biomedical BERT encoder. Returns an embedding of your text.
Entity-linking encoder (UMLS). Returns an embedding for a term/phrase.
Protein language model. Enter an amino-acid sequence to get a residue-pooled embedding.
Molecular transformer. Enter a SMILES string to get a molecular embedding.
Genomic foundation model. Enter a DNA sequence to get an embedding.
A biomedical CLIP model (PubMedBERT text encoder + ViT). Enter text to get a CLIP text embedding usable for zero-shot classification and retrieval.
DeepMind's Enformer predicts gene expression / regulatory signal from DNA. Enter a DNA sequence to get a sequence embedding. (Best with long sequences.)
AllenAI's scispaCy biomedical NER (en_core_sci_sm). Enter biomedical text to extract recognized entities.
Owkin's Phikon-v2 histopathology feature extractor (ViT). Upload an H&E tile to get a feature embedding for biomarker / tissue analysis.
Stanford's chest X-ray vision-language model. Upload a chest X-ray (and an optional question) to generate a description of the findings.
Segments 100+ anatomical structures from a CT volume. Upload a CT NIfTI (.nii.gz) to get the segmented structures and their relative volumes (fast mode).