HeadsUpAI

NVIDIA Launches Nemotron 3 Open Models for Agentic and Physical AI

· Updated

NVIDIA released new open models at GTC. Nemotron 3 Ultra delivers frontier-level intelligence with 5x throughput efficiency using the NVFP4 format on the Blackwell platform, powering coding assistants, search, and complex workflow automation. Nemotron 3 Omni integrates audio, vision, and language for AI agents extracting insights from videos and documents, while Nemotron 3 VoiceChat enables real-time conversations with simultaneous listening and response.

This expansion extends NVIDIA's open model strategy beyond language. Cosmos 3, expected soon, will unify synthetic world generation with physical AI reasoning for robotics. NVIDIA Isaac GR00T N1.7 is now commercially viable for humanoid robot deployment, and Proteina-Complexa accelerates protein drug discovery through BioNeMo, NVIDIA's open platform for healthcare and life sciences. Cursor, CrowdStrike, and Perplexity are already deploying across these model families.

Select models are available on GitHub and Hugging Face, with many also offered as NVIDIA NIM microservices for scalable deployment on any NVIDIA-accelerated infrastructure.

View the full update on nvidianews.nvidia.com
NVIDIA Newsroom
NVIDIA Newsroom
@nvidianewsroom
X

#NVIDIAGTC news: NVIDIA is expanding its open model families to power the next wave of agentic, physical and healthcare AI. New models across Nemotron, Cosmos and BioNeMo help developers build intelligent systems that reason and act across digital and physical worlds with partners like @LangChain and @GoogleDeepMind. 🔗 https://t.co/pyx98akJJ0

40retweets
View on X

Share this update