Nvidia Readies Trillion-Parameter Nemotron 4 to Challenge Open Models

Nvidia is expanding beyond the graphics processors that power generative AI into models and software that can deepen customer reliance on its technology stack. Its Nemotron family is designed to give enterprises and developers models they can adapt and deploy. A system exceeding one trillion parameters would signal a broader push to compete directly with leading open models while reinforcing demand for Nvidia’s computing and software ecosystem.

Nvidia is developing the Nemotron 4 family with a top-tier model containing at least one trillion parameters, according to the report, and could have it ready by late autumn. The company has not disclosed an investment amount. Nvidia also introduced Nemotron 3.5 Lightning and NeMo Switchyard, a model-routing library designed to select different models for different tasks, widening its open AI toolkit as competition among model providers intensifies.