Nvidia has introduced Nemotron 3 Ultra, a new open model with 550 billion parameters, expanding its lineup of AI systems for developers and enterprises.

The launch adds another large-scale option to Nvidia’s model family at a time when the company is positioning its software and infrastructure for so-called agentic AI, systems designed to carry out multi-step tasks with less human intervention. Nvidia did not frame the release as a consumer product, but as part of its broader effort to support enterprise and research use cases.

Nemotron 3 Ultra is one of the largest open models Nvidia has announced to date. The company described it as open, which means developers can inspect and build on the model more readily than with closed systems. That approach has become increasingly important in AI, where many organizations want more control over deployment, tuning and integration into their own platforms.

The release comes alongside a series of related Nvidia announcements focused on enterprise AI infrastructure and model acceleration. In recent blog posts, the company has highlighted partnerships and technical work tied to agentic systems, AI factories and accelerated computing. Those efforts suggest Nvidia is continuing to build a stack that spans chips, systems and models rather than relying on hardware alone.

Large models like Nemotron 3 Ultra are typically aimed at high-end workloads, including advanced reasoning, content generation and enterprise automation. Nvidia did not provide in the source material a detailed benchmark breakdown, pricing information or a full technical specification beyond the model size. Still, the scale alone places it among the largest open models in the market.

The company has been pushing more aggressively into model development in addition to its long-standing role as a maker of AI accelerators. That strategy reflects a broader industry trend, with major infrastructure providers increasingly offering their own models, frameworks and tooling to complement hardware sales. For customers, the appeal is often convenience and integration. For vendors, it helps deepen ties to the AI platforms businesses use.

Nvidia’s recent public messaging also points to growing interest in models that can work across the full lifecycle of enterprise AI, from experimentation to production deployment. The company has said agentic AI is moving from proof of concept into real-world use, and its latest announcements are framed around helping organizations build the systems needed for that transition.

Nemotron 3 Ultra arrives in that context as both a technical milestone and a strategic signal. By releasing a very large open model, Nvidia is reinforcing its role not just as a supplier of AI infrastructure but as an active participant in the model ecosystem itself.

The announcement did not include a release timeline for broader availability beyond the launch itself, and Nvidia has not yet detailed how Nemotron 3 Ultra will be distributed or supported across its platforms. Even so, the model’s size and open status are likely to draw attention from developers, researchers and enterprise teams looking for a new foundation for large-scale AI applications.