Overview
Use our open foundation models to build the next generation of biomedical AI, spanning electronic health records to medical imaging to genomics.Get Started
Quickstart
Set up your environment in minutesInstall dependencies, configure CUDA, and download your first model with our automated setup script or step-by-step guide.
Introduction
Learn about our approach and architecture
Models
Explore the full model catalog
Core Concepts
Understand JEPA and multimodal fusion
Model Families
SMB-v1-Structure
1.7B parameters · Biological World ModelOur flagship multimodal model that predicts patient trajectories using JEPA architecture. Integrates genomics, imaging, EHR, and proteomics.
SMB-EHR
4B parameters · Electronic Health RecordsFoundation model for clinical event prediction and temporal reasoning over patient records.
SMB-Vision
97M - 600M parameters · Medical ImagingVision encoders for radiology, pathology, and CT imaging tasks. Multiple sizes available.
SMB-Language
8B parameters · Clinical TextBiomedical language model for sentence similarity, semantic search, and text understanding.
All Available Models
| Model | Parameters | Type | Use Case |
|---|---|---|---|
| SMB-v1-1.7B-Structure | 1.7B | World Model | Patient trajectory prediction |
| smb-ehr-4b | 4B | EHR | Clinical event prediction |
| smb-vision-v0-risk | 0.6B | Vision | Risk assessment |
| smb-vision-v0-mim | 0.6B | Vision | Masked image modeling |
| smb-vision-large | 0.3B | Vision | General encoder |
| smb-vision-base | 97M | Vision | General encoder |
| smb-vision-ct-base-0519 | 97M | Vision | CT-specific |
| smb-mntp-llama-3.1-8b-v1 | 8B | Language | Sentence similarity |
Choose Your Model
I want to predict patient outcomes over time
I want to predict patient outcomes over time
Use Standard Model v1 — our JEPA-based world model that predicts how patients evolve given interventions.View Standard Model v1 documentation →
I'm working with electronic health records
I'm working with electronic health records
Use SMB-EHR-4B — trained on longitudinal clinical events for next-event prediction and temporal reasoning.View SMB-EHR documentation →
I need to analyze medical images
I need to analyze medical images
Start with SMB-Vision models. Start with
smb-vision-base for general tasks, or smb-vision-ct-base-0519 for CT scans.View SMB-Vision documentation →I'm doing clinical text analysis
I'm doing clinical text analysis
Use SMB-Language-8B for semantic search, sentence similarity, and clinical text embeddings.View SMB-Language documentation →
Hardware Requirements
| Model Size | GPU Memory | Examples |
|---|---|---|
| ~100M | 4-8 GB | smb-vision-base, smb-vision-ct-base |
| ~300M-600M | 8-16 GB | smb-vision-large, smb-vision-v0-risk |
| 1.7B | 16-32 GB | Standard Model v1 |
| 4B | 24-48 GB | smb-ehr-4b |
| 8B | 32-80 GB | smb-language |