NVIDIA Releases New Open AI Models for Drug Discovery, Robotics and Healthcare Research
NVIDIA introduced a suite of open-source AI models designed for healthcare researchers, roboticists and autonomous vehicle developers. The company released updated versions of its Nemotron, Cosmos, Isaac GR00T and BioNeMo model families, each tailored to specific industries.
For healthcare and life sciences, the focus is on accelerating drug discovery. NVIDIA's Proteina-Complexa model generates protein binders that bind to target proteins-a critical step in designing new therapeutics. Researchers at Novo Nordisk, Viva Biotech and Manifold Bio have already experimentally tested designs created by the model.
Protein Structure Database Expansion
NVIDIA collaborated with Google DeepMind, EMBL's European Bioinformatics Institute and Seoul National University to expand the AlphaFold Protein Structure Database. The partnership added 1.7 million high-confidence protein complex predictions from roughly 30 million calculated structures.
This dataset addresses a practical bottleneck in drug discovery: identifying which proteins interact with disease targets. Researchers can now explore more potential drug targets before moving to laboratory experiments.
Simulation Speed for Clinical Research
NVIDIA also released nvQSP, a GPU-accelerated simulation tool for pharmaceutical modeling. In tests, the tool delivered up to 77x faster performance than traditional CPU-based simulations, allowing researchers to model hundreds of dose levels and patient subpopulations in scenarios that previously took weeks.
This speed matters for preclinical work. Faster simulations let teams test more treatment scenarios before committing resources to clinical trials.
Platform and Access
The BioNeMo platform serves as the foundation for these healthcare models. It provides researchers with tools to model, design and simulate biological systems at scale.
NVIDIA is releasing these models as open-source software available on GitHub and Hugging Face. The company also offers them as NIM microservices-packaged software that can run on any NVIDIA-accelerated infrastructure, from edge devices to cloud systems.
Healthcare organizations can deploy these models on their own hardware without relying on cloud vendors, which addresses data privacy concerns common in pharmaceutical and clinical research.
Broader Model Releases
Beyond healthcare, NVIDIA released updates to its other model families. Nemotron 3 Ultra delivers what the company calls frontier-level performance with 5x throughput efficiency on NVIDIA's Blackwell processors. Nemotron 3 Omni integrates audio, vision and language understanding for AI agents that can extract insights from videos and documents.
For robotics, NVIDIA released Isaac GR00T N1.7, a vision-language-action model designed for humanoid robots. The company also previewed GR00T N2, expected later in 2026, which the company says succeeds at new tasks in unfamiliar environments more than twice as often as competing models.
Companies including LG Electronics, Humanoid and mimic robotics are adopting these models for robot deployment.
Learn more about AI for Healthcare and Generative AI and LLM technologies.
Your membership also unlocks:
