NVIDIA Expands Open Model Families for Agentic, Physical and Healthcare AI

NVIDIA has expanded its open model families to power the next wave of agentic, physical and healthcare AI, introducing specialized models that enable developers to build systems capable of reasoning and acting across digital and physical environments, according to an official announcement at GTC 2026.

The company's NVIDIA Nemotron family now includes omni-understanding models that deliver natural conversations, complex reasoning and advanced visual capabilities through multimodal integration. Nemotron 3 Ultra offers 5x throughput efficiency with the NVFP4 format on NVIDIA's Blackwell platform, targeting enterprise applications like coding assistants and workflow automation, while Nemotron 3 Omni integrates audio, vision and language understanding for video and document analysis, and Nemotron 3 VoiceChat enables real-time simultaneous listening and responding.

For physical AI applications, NVIDIA introduced NVIDIA Cosmos 3 as a world foundation model for generating synthetic environments to enhance robot and autonomous vehicle simulation, alongside NVIDIA Isaac GR00T N1.7 for robotics and NVIDIA Alpamayo 1.5 for autonomous vehicle reasoning with multi-camera support. These models address the critical need for AI systems to perceive and interact with physical environments, moving beyond language-based capabilities.

Healthcare applications are accelerated through the NVIDIA BioNeMo platform, featuring the Proteina-Complexa model for protein drug discovery, developed in collaboration with Google DeepMind, EMBL's European Bioinformatics Institute and Seoul National University. This includes a new open dataset of millions of AI-predicted protein complex structures, enabling faster biomedical research.

According to Kari Briski, vice president of generative AI software at NVIDIA, "Open source AI has become a global force for innovation. From biology and scientific discovery to robotics and autonomous machines, NVIDIA open model families extend intelligence beyond language, enabling developers worldwide to build intelligent agents and power breakthroughs across digital and physical industries."

Real-world adoption is already underway: CodeRabbit, CrowdStrike, Cursor, Factory, ServiceNow and Perplexity are deploying Nemotron for agentic AI; LG Electronics and Milestone Systems for physical AI; and Novo Nordisk, Viva Biotech and Manifold Bio for healthcare applications. The integration with LangChain enables businesses to build, deploy and monitor intelligent AI assistants for enterprise-scale task automation.

Edison Scientific is using NVIDIA Nemotron as a core component of Kosmos, an autonomous AI scientist that compresses months of research into a single day for over 50,000 researchers. This represents a significant shift from language-focused AI toward systems that can perform physical tasks and scientific discovery, with NVIDIA positioning open models as essential for "advancing innovation at global scale."

The expansion marks a strategic pivot from general-purpose language models to specialized frameworks for action-oriented AI. As noted in NVIDIA's documentation, this transition from "digital assistance to physical and clinical execution" requires precision where reasoning is coupled with domain-specific data to ensure reliability in real-world applications.

Industry analysts observe that NVIDIA's approach addresses a critical gap in AI development: while open models have become standard for language processing, the ability to build systems that interact with physical environments remains limited. By providing specialized foundation models for robotics, autonomous systems and biomedical research, NVIDIA is creating the infrastructure for AI to move from "talking" to "doing" across multiple industries.

NVIDIA's official announcement confirms these models are now available for developers, with the company emphasizing that "open model weights are the price of entry" but the real challenge lies in building secure, private orchestration layers for enterprise deployment.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn