Google Cloud Model Garden
Deploy and fine-tune biological foundation models on managed Vertex AI endpoints: protein structure, therapeutics, medical imaging, and pathology.
Overview
Google Cloud Model Garden is the model catalog inside Vertex AI, the place to deploy biological and health foundation models to managed inference endpoints on Google Cloud. Rather than shipping weights, it exposes models as one-click deployments and hosted APIs, so teams that need protein structure prediction, therapeutics prediction, or medical-imaging inference can stand up an endpoint without provisioning GPUs. It is the natural home for Google's Health AI Developer Foundations, extending the catalog well past the protein-only scope of most model hubs into pathology, dermatology, radiology, and health audio.
What you can run on Google Cloud Model Garden
The catalog hosts AlphaFold 2 for protein structure prediction alongside Google's Health AI Developer Foundations. MedGemma and TxGemma bring generative and predictive capabilities for clinical text and therapeutic development, and both can be fine-tuned on your own data. MedSigLIP provides medical image-text embeddings, while Path Foundation (digital pathology), Derm Foundation (dermatology), and CXR Foundation (chest radiography) supply domain-specific image encoders. HeAR, the Health Acoustic Representations model, covers health audio such as cough and respiratory sounds. Together these span protein structure, therapeutics, medical imaging, pathology, dermatology, radiology, and health-audio inference.
Inference and fine-tuning on Google Cloud Model Garden
Each model deploys as a managed Vertex AI endpoint through the Google Cloud console or Vertex AI APIs, giving you autoscaling hosted inference without managing hardware. MedGemma and TxGemma can be fine-tuned on your data before deployment, letting teams adapt them to specific clinical or therapeutic tasks. Everything runs under Google Cloud's standard enterprise controls for identity, networking, and audit logging, making it a fit for regulated healthcare and life-sciences organizations that need governed managed endpoints rather than raw weight downloads.
Run inference on Google Cloud Model Garden (8)
Protein structure prediction model that folds amino acid sequences into 3D structures with atomic accuracy, scoring a median GDT of 92.4 at CASP14.
Open medical multimodal models from Google, built on Gemma 3 with a medically tuned SigLIP vision encoder for clinical text and image understanding.
Medically tuned SigLIP encoder from Google that maps medical images and text into one embedding space for zero-shot classification and retrieval.
Chest X-ray embedding model built on ELIXR, producing image and image-text embeddings for data-efficient and zero-shot radiograph classification.
Open therapeutics foundation models from Google, built on Gemma-2, for drug-discovery property prediction and conversational reasoning.
Health acoustics foundation model that turns short clips of coughs and breaths into embeddings for building acoustic biomarker models with less data.
Histopathology foundation model that encodes 224x224 H&E patches into compact 384-dimensional embeddings for tumor and biomarker classifiers.
Google's dermatology image embedding model that produces 6144-dimensional embeddings for data-efficient skin-condition classifiers.
Fine-tune on Google Cloud Model Garden (2)
Open medical multimodal models from Google, built on Gemma 3 with a medically tuned SigLIP vision encoder for clinical text and image understanding.
Open therapeutics foundation models from Google, built on Gemma-2, for drug-discovery property prediction and conversational reasoning.