NVIDIA Launches Nemotron 3 Open Models to Power Agentic AI Systems
NVIDIA introduces the Nemotron 3 family of open models featuring Nano, Super, and Ultra sizes to help developers build efficient, transparent multi-agent AI applications. The new models utilize a breakthrough hybrid architecture to deliver leading accuracy and unmatched inference speeds.
NVIDIA unveils the Nemotron 3 family of open models, offering Nano, Super, and Ultra sizes tailored for building transparent and efficient agentic AI applications. This comprehensive release includes not just the models themselves, but also the training datasets and reinforcement learning libraries developers need to create highly accurate, specialized AI agents.
The new models feature a breakthrough hybrid latent mixture-of-experts architecture that directly tackles the challenges of modern multi-agent systems, such as communication overhead and high inference costs. Nemotron 3 Nano delivers four times higher throughput than its predecessor, providing the most tokens per second for scalable multi-agent deployments, while advanced reinforcement learning techniques ensure superior accuracy.
Major enterprise companies like Accenture, CrowdStrike, Oracle, and ServiceNow are already integrating these open models into their workflows across various industries. Furthermore, the Nemotron 3 family supports NVIDIA's sovereign AI initiatives by giving organizations worldwide the transparent, customizable tools required to build AI systems aligned with their specific data, regulations, and cultural values.