Foundation Models
Built for Scale. Engineered for Intelligence. Designed for Discovery.
Paige’s foundation models redefine what’s possible in pathology. Built on the world’s largest dataset of digitized slides, our technology captures the intricate complexities of tissue like never before—enabling new levels of discovery, precision, and performance.
Two Foundation Model Suites. Unlimited Possibilities.
Virchow: Vision AI
Virchow2
Expanded multi-scale training (H&E + IHC), boosting precision across 40+ tissue types
Virchow2G
Our largest model yet—1.8B parameters—optimized for high-stakes pathology AI
Virchow2G-Mini
A lightweight alternative, delivering high-throughput performance on limited compute
Virchow |
Virchow2 |
Virchow 2G |
Virchow 2G-Mini |
|
---|---|---|---|---|
Dataset | 1.5 million H&E WSIs | 3 million H&E & 400k IHC WSIs | 3 million H&E & 400k IHC WSIs | 3 million H&E & 400k IHC WSIs |
Tissue Types | 17 tissue types | 40 tissue types | 40 tissue types | 40 tissue types |
Parameters | 632 million parameters | 632 million parameters | 1.8 billion parameters | 21.6 million parameters |
Magnifications | 20X magnifications | 5x, 10x, 20x, 40x magnifications | 5x, 10x, 20x, 40x magnifications | 5x, 10x, 20x, 40x magnifications |
Publication | Published in Nature Medicine [paper link] | Arxiv Paper | Arxiv Paper | Arxiv Paper |
Availability | Hugging Face + Azure AI Foundry | Hugging Face + Azure AI Foundry + Commercial License | Commercial License Only | Commercial License Only |
PRISM: Multi-Modal AI
Combines Large Vision Models (LVMs) with Large Language Models (LLMs) for slide-level analysis.
Built using Virchow, PRISM analyzes and extracts insights from tissue at the whole slide level
Automated Clinical Report Generation | Converts whole-slide images into free-text pathology reports
Zero-Shot Cancer Classification | Detects cancer subtypes and predicts biomarkers without extensive labeled data
Multi-Modal Intelligence | Trained from 587K slides and 195K clinical reports to generate text-based insights
PRISM |
---|
587k H&E WSIs |
16 tissue types |
195,000 clinical text reports |
20x magnification |
Automatic clinical report generation complete with cancer detection, subtyping, and biomarker predictions |
Arxiv Paper |
Hugging Face + Azure AI Foundry |
AI That Works for You
Faster AI Model Development
Reduce the time and cost of training from scratch
Regulatory-Grade Performance
Built with clinical excellence and compliance in mind
Available Integration
Compatible with leading digital pathology platforms
