NVIDIA has announced the new NVIDIA Nemotron™ 3 family, which includes open models, data, and libraries to help build efficient and specialized AI systems across different industries.
The Nemotron 3 models come in three sizes: Nano, Super, and Ultra. These models use a new hybrid mixture-of-experts (MoE) architecture, making it easier for developers to build reliable, large-scale AI systems that work together.
As businesses move from simple chatbots to AI systems that collaborate in teams, developers are facing new challenges like managing communication, staying on track, and keeping costs low. Plus, they need transparency to trust the AI models that will handle important tasks. The Nemotron 3 models cater to these challenges, offering high performance as well as openness so that companies can create AI that fits their specific needs.
The Nemotron 3 models are part of NVIDIA's bigger push for sovereign AI, where organizations can build AI systems that align with their data, rules, and values. Many companies, including Accenture, Oracle, Siemens, Zoom, and more, are already using these models to improve their AI systems for fields like manufacturing, cybersecurity, and software development.
As AI systems become more complex, developers are using both proprietary models for advanced reasoning and open models like Nemotron 3 to save on costs. These open models allow companies to manage cutting-edge AI with efficiency.
The Nemotron 3 models are also ideal for startups looking to quickly develop and deploy AI systems. For example, companies backed by General Catalyst and Mayfield are using Nemotron 3 to create AI teammates that can work alongside humans.
Executive Statement
According to Jensen Huang, founder and CEO of NVIDIA, open innovation is the foundation of AI progress. With Nemotron, they are revolutionizing advanced AI into an open platform that gives developers the transparency and efficiency they need to build agentic systems at scale.