Building the World Model of Biology

For research, for trials, for life.

Biology studied in silos stays incomplete. We train foundation models natively across modalities and scales — learning the dynamics of human biology to predict what happens next.
Discover M-OptimusRead our Research
Trusted by
16 of the top 20 pharma companies
|
1,000+ institutions worldwide

Built to learn across every scale, from day one

Multimodal

Multimodal. Multi-scale. By design.

The cross-modal patterns that drive our response to drugs are invisible when data are analyzed in isolation. We built our models from day one to jointly learn from many scales, from cell to tissue to organ. That's what it takes to explain why two patients with the same clinical profile respond differently to the same drug.
Data Engine

Generating biological data. At scale.

The biological data that matters most doesn't yet exist at scale; it has to be generated. STELA is our multi-institutional initiative to do exactly that: deeply profiled, clinically linked, multimodal patient data spanning geographies and diseases. Patients are profiled across modalities, so that every person in the atlas improves care for patients they will never meet.
Team

World-class research. In ML and biology.

Most labs have one or the other. Ours has both: frontier ML researchers and deep biological expertise from Google Brain, Google DeepMind, Meta, Owkin, MIT, ETH Zürich, Cambridge, Helmholtz, and more. 60% of our team hold PhDs. Guided by a scientific and strategic advisory board at the frontier of biology and AI.

Get more from your existing data

Starting with the histology slides, sequencing data and clinical records already in your archives.
Drug target & biomarker discovery
01

Predict spatial distribution of ADC drug targets, immune microenvironment, and stromal architecture from routine histology slides. No new assays required to start.

Indication expansion
02

Your drug works — but in which patients, and in which other diseases? Our models score archival cohorts across disease, stage, and mechanism of action, identifying patient subsets with the spatial biology your drug needs to succeed. Turn thousands of archived H&E slides from completed trials into a biological screening engine.

Diagnostics & spatial biology
03

Integrate signals across histology, spatial transcriptomics, and genomics from a single pipeline. Clinically actionable biology at the resolution the standard workup doesn't reach.

Treatment response & trial design
04

Two patients. Same diagnosis. Same standard biomarkers. Different spatial biology — and different outcomes. Our models discover multimodal biomarker signatures that distinguish responders from non-responders, even from small cohorts in early-phase trials. Designed to increase the probability of success before Phase III and asset rNPV.

The standard for excellence: #1 in independent benchmarks

Validated by independent teams and peer-reviewed research, our models consistently hold the #1 rank across industry-standard public benchmarks.
H-Optimus-1
6.06
Virchow2
6.34
H-Optimus-0
6.86
UNI2
7.10
mSTAR
7.65
Pathology Model - Benchmark Ratings

Chart shows overall rank across all tasks, PathBench (lower is better)

#1

Ranked Model

100+

Papers published using our models

16/20

Top 20 pharma companies

1000+

Clinical Practices

1,184,533

Total Model Downloads

Biology, decoded at every scale

While our histology model, H-Optimus, set the standard, M-Optimus goes further, integrating pathology, spatial transcriptomics, and genomics into a single architecture. It is the first foundation model built to understand biology across every scale, delivering even deeper insights and more powerful performance than was possible before.

Request access

Proven in Practice: Powering Science at the World's Leading Institutions

Case Study
-
February 3, 2026
Breast Cancer Recurrence Risk Prediction with H‑Optimus‑1 and STAMP
Read case study
Case Study
-
December 19, 2025
ICGI researchers build a winning pathology report generation model with H-optimus-1
Read case study
Case Study
-
December 11, 2025
H-Optimus-1 for streamlining Spatial Transcriptomics workflows in IBD research
Read case study
Case Study
-
November 3, 2025
MIT partners with Bioptimus for Clinical Cancer Prediction Study
Read case study
Subscribe to our Newsletter

Contribute data.
Benefit from the atlas.

STELA is our collaborative data engine, a growing, multi-institutional network of deeply profiled patients. Build with us.
Become a Partner

Use our models.
See what your data can reveal.

Start integrating multimodal foundation models into your development pipeline with M-Optimus or access the benchmark-setting pathology model, H-Optimus.
Get Started

Selected Partnerships