About
GenBio AI is building AIDO, a world model for human biology: a programmable virtual cell and multiscale foundation-model system that predicts, generates, and simulates biology from molecules to phenotypes. It targets researchers and professionals in pharmaceutical, healthcare, biotechnology, and life-sciences fields; its differentiator is an integrated end-to-end model spanning DNA, RNA, protein, cells, and higher-level biological behavior rather than isolated components.
Market
GenBio AI competes in AI-native computational biology, biological foundation models, and virtual-cell platforms serving drug discovery, bioengineering, and broader life-science research. Its positioning is unusually broad and infrastructure-oriented: AIDO aims to connect biology from molecules to phenotypes, while VCHarness autonomously creates, tests, and refines executable models rather than merely applying a fixed model. This contrasts with competitors that emphasize virtual-cell drug-discovery simulations, cancer and clinical-outcome modeling, multi-omics pathway discovery, or virtualized biological experiments.
GenBio AI primarily targets R&D organizations in pharmaceuticals, biotechnology, and life sciences, including computational biology, drug-discovery, genomics, oncology, and personalized-medicine teams. Its users and buyers are likely technical research leaders, computational biologists, and drug-discovery scientists who need configurable models, virtual-cell simulations, and faster hypothesis testing; the sources do not specify a precise company-size segment.
At a Glance
Problem
Drug discovery and biological research are constrained by the complexity of biology and by fragmented tools that model only one biological scale at a time. GenBio AI characterizes the problem as a research-silo issue: molecular, cellular, and organism-level data are difficult to combine into a coherent model of living systems. The economics are severe, with the company citing traditional drug development costs of $1–2 billion and a 90% failure rate. Its clearest use case is reducing the time and uncertainty involved in testing biological hypotheses, such as predicting how cells respond to CRISPR edits or potential therapies.
The underlying pain is not just computation but experimentation: building a useful virtual-cell model can require months of expert iteration without guaranteeing a useful result. GenBio AI aims to turn that process into a faster simulation-and-learning loop, allowing researchers to test genetic, chemical, or environmental interventions in silico before committing as heavily to laboratory work.
Product / Service
GenBio AI is developing AIDO, an AI-driven digital organism and world model for human biology. It combines multiscale foundation models covering DNA, RNA, proteins, protein structure, single-cell expression, and evolution into an integrated system that can predict, generate, and simulate biological processes. The initial AIDO platform is an interactive toolkit for building models across molecules, cells, and tissues, adapting them to sparse data, and customizing and visualizing results for research questions.
The current delivery model is an openly distributed research-software ecosystem rather than a clearly documented commercial SaaS product. AIDO.ModelGenerator lets scientists adapt, fuse, fine-tune, and benchmark GenBio AI and open-source models using their own data, with releases through public repositories. Its newer VCHarness system adds an autonomous agent loop that proposes models, writes code, runs experiments, evaluates results, and iterates; in a CRISPR-response benchmark, GenBio AI says it found models that outperformed expert-designed baselines in days rather than months. The intended benefit is a more integrated and efficient workflow for drug design, target identification, mRNA-vaccine optimization, disease modeling, and synthetic biology.
Market
GenBio AI competes in biological foundation models, generative biology, AI-enabled drug discovery, and the emerging virtual-cell or biological-world-model category. Its differentiation is the attempt to connect multiple biological modalities and scales into one programmable system, rather than offering a model for only proteins, genomes, cells, or images. Publicly identified alternatives include Generate Biomedicines, Profluent Bio, and Bioptimus; in adjacent foundation-model work, GenBio AI also benchmarks against models such as UNI2-H and H-Optimus-1.
The company has clear technical and ecosystem traction: it released Phase 1 of AIDO with six models in December 2024, released AIDO.ModelGenerator and public model resources in 2025, announced a June 2026 collaboration with NVIDIA to scale virtual-cell development, and reports state-of-the-art results for its open-weight GenBio-PathFM model. Commercial traction is less transparent. Tracxn reports that GenBio AI has raised no funding, while Latka estimates $4.5 million of 2025 revenue and a $13.5 million valuation; because those third-party signals conflict and the gathered evidence does not establish official customer or revenue disclosures, the company is best described as an early-stage, technically active platform with unverified commercial scale rather than definitively pre-revenue.
Founders & Leadership
Recent News
GenBio AI announced a collaboration with NVIDIA to accelerate virtual-cell world models, integrating NVIDIA BioNeMo, Megatron, NIM microservices, and BioNeMo Agent technologies. The effort targets AI systems that simulate human cellular behavior across biological modalities and scales.
The Innovator profiled GenBio AI as a startup combining virtual-cell world models and autonomous model-building technology with NVIDIA accelerated computing and the BioNeMo ecosystem.
GenBio AI proposed an operational definition of a virtual cell based on AI world-model architecture, emphasizing action-conditioned simulation, counterfactual reasoning, and long-horizon planning.
GenBio AI introduced VCHarness, a closed-loop system combining biological foundation models, an AI coding agent, and search to autonomously build and improve perturbation-response models. The company reported that it performed at or near the top across four cell lines and discovered new model designs.
GenBio AI described AIDO.Tissue, a spatial-transcriptomics model that uses multi-cell neighborhoods rather than isolated cells to capture intra- and inter-cellular dependencies and interpret what a cell means in context.
GenBio AI introduced GenBio-PathFM, a 1.1-billion-parameter histopathology foundation model. The company reported state-of-the-art results on several public benchmarks while using 10–20% of the data required by leading models, and described it as a strong open-weight model trained exclusively on public data.
GenBio AI evaluated more than 600 model variants for perturbation-response prediction and reported that results for Jurkat and K562 cell lines reached the estimated experimental error limit. The findings suggest that foundation models can produce highly accurate cellular simulations when trained with sufficient data and the appropriate modality.
This GenBio AI research post reviewed large-model approaches for single-cell omics and drug discovery, including how models such as scGPT learn from large public datasets and transfer to annotation, mapping, and perturbation-prediction tasks.
GenBio AI presented three sparse mixture-of-experts AIDO.DNA2 variants with 235 million, 470 million, and 1 billion total parameters. The company reported improved ClinVar variant-effect prediction over earlier AIDO.DNA versions and the similarly scaled Evo2-1B baseline.
The proposed approach separates a cell’s baseline state from the effect of a perturbation, with the goal of improving interpretability and reliability while identifying signals caused by batch effects, rare cell types, or weak measurements.
Active Roles
3Business Model
The best-supported model is partnership-led enterprise commercialization: GenBio AI develops foundation models for drug design, bioengineering, and personalized medicine, makes model resources available through Hugging Face and GitHub, and solicits partnerships for commercial applications. Public evidence does not disclose pricing or a confirmed subscription, API, licensing, or usage-based revenue stream; the site's stated license is for non-commercial informational use.
Products
Tech Stack
Similar Companies
Competitors
Key Investors
not found in sources