NVIDIA Expands Open Models Across Three Key Domains
NVIDIA announced a major expansion of its open model families at GTC 2026. The company introduced new models that help developers and scientists build intelligent systems. These systems can reason and act across digital and real-world environments. Furthermore, the expansion targets three critical verticals: agentic AI, physical AI, and healthcare.
Open models drive innovation at global scale. NVIDIA’s growing portfolio includes Nemotron for agentic systems, Cosmos for physical AI, Alpamayo for autonomous vehicles, Isaac GR00T for robotics, and BioNeMo for biomedical research. Together, these frameworks unlock new capabilities across industries.
“Open source AI has become a global force for innovation,” said Kari Briski, VP of generative AI software at NVIDIA. “NVIDIA open model families extend intelligence beyond language, enabling developers worldwide to build intelligent agents and power breakthroughs.”
Nemotron 3 Models Power Agentic AI
Four New Models Extend Multimodal Intelligence
The NVIDIA Nemotron family now includes omni-understanding models across language, vision, voice, and safety. These models help developers build specialized agentic AI. Each model serves a distinct purpose.
- Nemotron 3 Ultra delivers frontier-level intelligence with 5x throughput efficiency using NVFP4 format on the Blackwell platform. Developers use it for coding assistants, search, and complex workflow automation.
- Nemotron 3 Omni integrates audio, vision, and language understanding. It allows AI agents to extract insights from videos and documents with high efficiency.
- Nemotron 3 VoiceChat supports real-time conversations. The model combines speech recognition, large language model processing, and text-to-speech in one system.
- Nemotron safety models detect unsafe content across text and images. Additionally, an agentic retrieval pipeline improves the relevance and accuracy of outputs.
Industry Adoption Is Rapidly Growing
LangChain has integrated Nemotron models into its agent development platform. As a result, businesses can build and monitor intelligent AI assistants that automate complex tasks. Companies such as CrowdStrike, ServiceNow, Cursor, Factory, and Perplexity deploy Nemotron models for advanced agentic applications.
Moreover, Edison Scientific uses Nemotron as a core component of Kosmos, an autonomous AI scientist. Kosmos serves more than 50,000 researchers and compresses months of research into a single day.
NVIDIA also released Nemotron-Personas, a collection of privacy-preserving synthetic datasets built from local census data. These datasets support sovereign AI development across multiple countries.
New Models Advance Physical AI Reasoning
Robots and Vehicles Gain Stronger AI Foundations
NVIDIA accelerates autonomous system development with new foundation models and simulation tools. These tools help robots and vehicles perceive, reason, and act in the physical world.
- NVIDIA Cosmos 3 is the first world foundation model to unify synthetic world generation, physical AI reasoning, and action simulation. It helps physical AI operate in complex environments.
- NVIDIA Isaac GR00T N1.7 is an open reasoning vision-language-action (VLA) model built for humanoids. It is now commercially viable for real-world deployment.
- NVIDIA Alpamayo 1.5 supercharges autonomous vehicle reasoning with navigation guidance, prompt conditioning, and flexible multi-camera support.
GR00T N2 Sets New Benchmarks
Jensen Huang previewed GR00T N2, a next-generation robot foundation model, during the GTC keynote. This model helps robots succeed at new tasks more than twice as often as leading VLA models. Furthermore, GR00T N2 currently ranks No. 1 on MolmoSpaces and RoboArena for generalist robot policies. NVIDIA expects the model to arrive before year-end.
Companies including Johnson & Johnson MedTech, LG Electronics, and Toyota Research Institute already use NVIDIA Cosmos and Isaac GR00T N1.7 to scale physical AI training.
Open Models Accelerate Healthcare Research
BioNeMo Platform Drives Drug Discovery
NVIDIA advances AI-driven discovery through open, multimodal foundation models for healthcare. The BioNeMo platform enables researchers to model and simulate biological systems at scale.
Proteina-Complexa is a generative model for protein binder design. It accelerates structure-based drug discovery and therapeutic development. Notably, Novo Nordisk, Viva Biotech, and Manifold Bio use this model to design proteins that bind to target proteins. All three companies have experimentally tested the generated designs.
AlphaFold Database Expands Significantly
NVIDIA collaborated with EMBL’s European Bioinformatics Institute, Google DeepMind, and Seoul National University to expand the AlphaFold Protein Structure Database. Together, they calculated approximately 30 million protein complex predictions and added 1.7 million high-confidence entries. Consequently, researchers can discover new drug targets and disease biology faster than before.
Additionally, NVIDIA introduced nvQSP, a GPU-accelerated simulation engine for pharmaceutical researchers. In benchmark tests, nvQSP delivered up to 77x faster performance compared with traditional CPU simulations. As a result, scientists can analyze hundreds of dose levels in the time it previously took to simulate just a few.
Availability and Deployment
Select NVIDIA open models, data, and frameworks are available on GitHub and Hugging Face. They also appear across cloud, inference, and AI infrastructure platforms, as well as on build.nvidia.com. In addition, many models are available as NVIDIA NIM microservices, enabling secure and scalable deployment on any NVIDIA-accelerated infrastructure.
