Nvidia launched Nemotron 3 Super, a 120-billion-parameter open artificial intelligence model designed to power complex agentic AI systems at scale.

The model combines advanced reasoning and high efficiency to help autonomous AI agents complete tasks more accurately while lowering computing costs, supported by a "1-million-token context window" that allows agents to retain large workflow data and improve performance in complex, multi-step tasks, the company said Wednesday in a blog post.

Technology companies are integrating the model into AI development tools, while enterprise software firms are deploying it to automate workflows in industries including telecommunications, cybersecurity and semiconductor design, Nvidia said.

The model, which uses a "hybrid mixture-of-experts" architecture to improve speed and accuracy, is available through Nvidia's developer platform as well as services such as Hugging Face and OpenRouter, the company said.