What Bioptimus does

Bioptimus is a Paris-based AI startup founded in 2024 that builds multimodal foundation models for biology, targeting pharmaceutical R&D, translational research, and precision medicine. The company's product stack comprises three pillars: H-Optimus, a 1.1 billion-parameter Vision Transformer pathology foundation model (H-Optimus-0 open-source under Apache 2.0; H-Optimus-1 ranked #1 on PathBench and HEST benchmarks across 229 tasks); M-Optimus, the first foundation model that jointly learns from H&E pathology, bulk RNA-seq, spatial transcriptomics, and clinical data to predict the activity of 20,000 genes from routine slides; and STELA, a multi-institutional data engine with 10x Genomics and Broad Clinical Labs targeting up to 100,000 profiled patient specimens across the US, Europe, and Asia. Underlying architecture combines ViT-g/14 image encoders with multimodal pretraining, supporting applications including drug target discovery, biomarker identification, indication expansion, and clinical trial enrichment.

Name

Bioptimus

Legal name

Bioptimus

Website

https://www.bioptimus.com

Company type

Private

Founded year

2024

Operating status

Operating

Headcount range

11–50 employees

Short description

Bioptimus builds multimodal foundation models for biology (H-Optimus, M-Optimus, STELA) used by 16 of the top 20 pharma companies and 1,000+ academic institutions for drug discovery, biomarker development, and clinical trial enrichment.

Ownership category

akta.pro rank

Product category

AI Foundation Models for Biology

NAICS

Research and Development in Biotechnology (except Nanobiotechnology) (541714), Software Publishers (51321)

SIC

Services-Commercial Physical & Biological Research (8731)

akta.pro primary industry

Bioinformatics, Multi-Omics Data Analysis & Interpretation Services (HLAGANAI)

akta.pro secondary industries

Bioinformatics & Multi-omics Analysis Software (HLAGAJAD), AI/ML Platforms for Drug Discovery & Experiment Optimization (HLAGAJAO)

Keywords

• Foundation models biology
• Computational pathology AI
• Multimodal biomedical models
• Drug discovery AI
• Spatial transcriptomics

Headquarters

HQ city

Paris

HQ country

France

HQ region

Europe

Offices1 record

Markets served

GTM type

B2B

Offering type

Software

Cost components

Personnel, Technology or R&D, Infrastructure, Operations, Marketing or Sales

Revenue model

1. Industry / commercial licensing: Commercial use of H-Optimus-1, H-Optimus-0, M-Optimus, and STELA data atlas requires a separate licensing agreement (industry licenses). Deployed through Amazon SageMaker on AWS Marketplace for enterprise customers. Used by 16 of top 20 pharma companies.
2. Academic / non-commercial access: Universities and research institutions can request non-commercial research licenses for evaluation and experimental use. H-Optimus-0 is open-source under Apache 2.0; H-Optimus-1 and H0-mini are under CC-BY-NC-ND 4.0 for academic use.
3. Strategic pharma collaborations & services: Co-development and pilot deployments with pharmaceutical partners (top pharma already onboarded as pioneer partners for M-Optimus), including access to STELA data backbone for joint drug discovery, clinical trial enrichment, and biomarker development programs.

Pricing tiers

Go-to-market motion1 record

Distribution channels4 records

Marketing channels9 records

Core offering

Bioptimus builds and licenses AI foundation models for biology. Its core products are H-Optimus (a state-of-the-art pathology foundation model trained on over 1 million H&E whole-slide images), M-Optimus (a multimodal world model that jointly learns from histology, RNA-seq, spatial transcriptomics, and clinical data), and STELA (a multi-institutional spatial biology atlas initiative). Models are licensed commercially to pharmaceutical and biotechnology customers (with academic/non-commercial access via Hugging Face and AWS Marketplace).

Product overview

Bioptimus offers a unified AI foundation model platform for biology structured around three integrated pillars: (1) H-Optimus, a single-modality pathology foundation model family (H-Optimus-0, H-Optimus-1, H0-mini) that serves as the histopathology backbone for biomarker discovery, patient stratification, and clinical decision support; (2) M-Optimus, the multimodal multi-scale 'world model of biology' that extends H-Optimus by jointly learning from H&E, bulk RNA-seq, spatial transcriptomics, and clinical data to predict gene expression and trial outcomes; and (3) STELA, a collaborative data generation engine that feeds M-Optimus with deeply profiled, clinically linked multimodal patient specimens. Together, these products form a vertically integrated stack in which H-Optimus provides the visual foundation, M-Optimus the reasoning and prediction layer, and STELA the proprietary data backbone, all designed to accelerate drug discovery, de-risk clinical trials, and advance precision medicine.

Differentiator

Problem solved

Functional benefit

Brands

• H-Optimus: Bioptimus's pathology foundation model family (H-Optimus-0 and H-Optimus-1) for digital histopathology, providing state-of-the-art performance across pathology benchmarks.
• M-Optimus
• STELA

Products and services

• H-Optimus-1 A state-of-the-art 1.1 billion parameter Vision Transformer (ViT-g/14) foundation model for digital pathology, trained on over 1 million H&E whole-slide images from more than 800,000 patients across 4,000+ clinical centers and 50+ organs. Ranked #1 on PathBench across 229 tasks; available on AWS Marketplace and Hugging Face for academic use. Sold via enterprise licensing to pharmaceutical and biotech customers.
• H-Optimus-0 The original open-source 1.1 billion parameter pathology foundation model, trained on over 500,000 histology slides from 4,000+ clinical practices. Released under Apache 2.0 license, designed for broad academic and research use via Hugging Face.
• H0-mini A lightweight, computationally efficient pathology model distilled from H-Optimus-0 in collaboration with Owkin, delivering comparable performance to larger foundation models at significantly reduced inference cost. Released under CC-BY-NC-ND 4.0 license for non-commercial academic research on Hugging Face.
• M-Optimus A multimodal, multi-scale foundation model that combines H&E pathology images, bulk RNA-seq, spatial transcriptomics, and clinical data into a unified embedding space. The first model built to understand biology across every scale, enabling digital twin simulations, drug target prediction, biomarker discovery, and trial design optimization. Request-access only via direct enterprise license.
• STELA (Spatial Tissue Embedding Learning Atlas) A multi-institutional collaborative data initiative, launched with 10x Genomics and Broad Clinical Labs, designed to generate the world's largest clinically linked multimodal spatial biology atlas profiling up to 100,000 patient tissue specimens across the US, Europe, and Asia. Serves as the proprietary data backbone for M-Optimus.

Quantifiable outcome

• H-Optimus-1 ranked #1 on PathBench with score 6.06 (vs Virchow 26.34, H-Optimus-0 6.86, UNI 27.10, mSTAR 7.65)
• +11 more outcomes

Named customers7 records

Segments4 records

Ideal customer profiles3 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

Integration5 records

AI capability10 records

Feature7 records

Partnerships

14 partnerships are on record, tiered flagship and minor.

• 10x GenomicsflagshipStrategic or Co-development Partner · 25 March 2026Co-launch partner of the STELA multinational spatial biology atlas initiative announced March 25, 2026. Provides the Xenium spatial transcriptomics platform used to profile up to 100,000 patient specimens across US, Europe and Asia. Bioptimus plans to expand from Xenium to Atera in 2027 as part of STELA. Michael Schnall-Levin (CTO and Founding Scientist of 10x Genomics) is a Bioptimus Strategic Advisor.
• Broad Clinical LabsflagshipStrategic or Co-development Partner · 25 March 2026Co-launch partner of the STELA initiative alongside Bioptimus and 10x Genomics. Contributes high-throughput spatial data generation capabilities to profile up to 100,000 patient tissue specimens across three continents.
• Massachusetts Institute of Technology (MIT)flagshipStrategic or Co-development Partner · 3 November 2025MIT partners with Bioptimus for the ARPA-H ADAPT Clinical Cancer Prediction Study (case study published November 3, 2025). MIT team led by Professor Regina Barzilay validated H-Optimus-1 as top-performing pathology foundation model for predicting Progression-Free Survival in metastatic breast, lung, and colon cancer. Regina Barzilay (MIT Jameel Clinic) also serves on Bioptimus's Scientific Advisory Board.
• NVIDIAflagshipTechnology or Integration · 11 June 2025Bioptimus secured access to NVIDIA compute resources (via DGX Cloud Lepton) in June 2025 to accelerate training of its foundation models for biology, in collaboration with Sofinnova Partners.
• ProsciaminorStrategic or Co-development Partner · 23 October 2024Collaboration announced October 23, 2024 to accelerate AI innovation in scientific research and drug development for improved patient care; jointly hosted a December 2024 webinar on foundation models in pathology AI.
• OwkinflagshipStrategic or Co-development Partner · 20 February 2024Founding-partner relationship — Bioptimus co-founders include ex-Owkin scientists (Jean-Philippe Vert was Chief R&D Officer; David Cahané was Chief Solutions Officer; Charlie Saillard worked at Owkin for 6 years; Philip Conti-Permanne was Director Corporate Development). Provides multimodal patient data, expertise, and a co-development collaboration for the CytoSyn diffusion model (released Nov 2025 on Hugging Face). Thomas Clozel (Owkin co-founder/CEO) serves as a Strategic Advisor.
• Amazon Web Services (AWS)flagshipTechnology or Integration · 20 February 2024Provides compute resources and cloud infrastructure. H-Optimus-1 is deployed on AWS Marketplace via Amazon SageMaker (launched May 2025) and M-Optimus / H-Optimus-1 are made available via Amazon SageMaker AI partnership.
• University of Cambridge / Wellcome Sanger Institute (Sarah Teichmann)minorStrategic or Co-development PartnerSarah Teichmann (Chair in Stem Cell Medicine, University of Cambridge; co-founder EnsoCell Therapeutics) serves as Chair of Bioptimus's Scientific Advisory Board.
• Columbia University / Chan Zuckerberg Biohub New York (Andrea Califano)minorStrategic or Co-development PartnerAndrea Califano (President of CZI Biohub New York; Clyde and Helen Wu Professor of Chemical Systems Biology at Columbia) sits on Bioptimus's Scientific Advisory Board.
• Broad Institute / MIT (Caroline Uhler)minorStrategic or Co-development PartnerCaroline Uhler (Director of the Eric & Wendy Schmidt Center at the Broad Institute; Professor of Engineering at MIT) is on Bioptimus's Scientific Advisory Board.
• Technical University (TU) Dresden (Jakob Kather)minorStrategic or Co-development PartnerJakob Kather (Professor for Clinical AI at TU Dresden; senior physician in medical oncology) is on Bioptimus's Scientific Advisory Board. The Kather Lab at TU Dresden is also a Bioptimus case-study customer using H-Optimus-1 for breast cancer recurrence risk prediction.
• Parker Institute for Cancer Immunotherapy (PICI) (John Connolly)minorStrategic or Co-development PartnerJohn Connolly (CSO at PICI) sits on Bioptimus's Scientific Advisory Board.
• Berlin Institute for Medical Systems Biology / Max Delbrück Center (Nikolaus Rajewsky)minorStrategic or Co-development PartnerNikolaus Rajewsky (Scientific and Founding Director, BIMSB at MDC) is on Bioptimus's Scientific Advisory Board.
• Gustave Roussy / ESMO (Fabrice André)minorStrategic or Co-development PartnerFabrice André (President of ESMO; Director of Research at Gustave Roussy) is on Bioptimus's Scientific Advisory Board.

Scale indicators11 records

Recent moves6 records

Expansion highlights7 records

Emerging players

• Insitro: Insitro uses ML on multimodal biological data (imaging, genomics, functional assays) for drug discovery. It overlaps with Bioptimus on the ML-meets-biology thesis and pharma partnership model, but focuses on owned therapeutic pipelines rather than platform licensing.
• Exscientia: Exscientia applies AI to small-molecule drug discovery, with a precision medicine focus and pharma partnerships. It is thematically adjacent to Bioptimus in AI-driven drug discovery but operates in chemistry-first rather than multimodal-biology-first territory.
• Recursion Pharmaceuticals: Recursion combines high-throughput biology with AI foundation models to discover drugs. It is an emerging player in the AI-for-biology category, with comparable multimodal data ambitions (phenoomics, transcriptomics) and a parallel pharma partnership and pipeline model.

Direct peers

• Owkin: Owkin is Bioptimus's founding partner and most direct competitor, building AI for patient stratification, drug discovery, and pathology. The two share scientific DNA (multiple Bioptimus co-founders came from Owkin) and overlapping multimodal patient data, but Owkin has deeper pharma deployment (e.g., Roche alliance) and a longer commercial track record.
• Paige.AI: Paige builds AI foundation models and clinical decision-support tools for digital pathology, including the FDA-authorized Paige Prostate. It is a direct competitor in pathology foundation models targeting pharma R&D and clinical diagnostics, with comparable multimodal ambitions and a similar customer base in oncology.
• DeepHealth (a RadNet company): DeepHealth applies AI to digital pathology and diagnostic imaging, including its iCAD-acquired and internally developed pathology models. As a SaMD-focused pathology AI company with FDA clearances, it represents Bioptimus's most direct competitor on the clinical commercialization path.
• PathAI: PathAI develops AI-powered pathology tools for both pharma research (biomarker discovery, clinical trial support) and clinical diagnostics. It is a direct peer in computational pathology foundation models, with a strong pharma partnership model and overlapping multimodal ambitions.

Broad incumbents

• Tempus Labs: Tempus is a broad incumbent in AI-driven precision medicine, combining multimodal patient data (pathology, genomics, clinical) with pharma R&D services. While not a foundation-model-first company, it is a key competitor for pharma AI budgets and multimodal data partnerships in oncology.
• Isomorphic Labs: Isomorphic Labs is Alphabet's AI-for-drug-discovery company, building foundation models of biology (AlphaFold-adjacent). It is a broad incumbent in the same category as Bioptimus's 'world model of biology' ambition, with vastly greater compute resources and a strategic pharma partnership agenda.

Regional players

• Artera: Artera (the 10x Genomics Atera platform partner) is a digital pathology and multimodal AI company developing prostate cancer and other clinical AI tests. As a 10x Genomics ecosystem partner overlapping with STELA, it is a regional/niche peer in AI-driven pathology biomarker development.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat5 records

Key risks6 records

Key highlights7 records

Customer concentration

• YouTubeyoutube.com/@bioptimus
• LinkedInlinkedin.com/company/bioptimus
• Xx.com/bioptimus_ai

Compliance1 record

Revenue estimate

Valuation estimate

Number of profiles

Profiles10 records

Funding overview

Funding rounds2 records

Investors15 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Bioptimus