Nvidia is building a huge open-source AI model called Nemotron 4

Nvidia is working on a new AI model series called Nemotron 4, according to people involved. The US chip giant wants to compete with the best open-source models out there. Nvidia is one of the few big US tech companies that actively releases open-source models. As AI development and running costs go up, cheaper Chinese models are getting close to the top US AI labs in performance, so open-source models are getting more attention. Nvidia hopes to expand AI use through an open model ecosystem and boost demand for its GPUs. Kari Briski, VP of generative AI, said they invest in Nemotron because they believe every company and country needs accessible frontier open-source models. The biggest Nemotron 4 model is expected to have at least one trillion parameters and might be ready by late fall. Nvidia also launched Nemotron 3.5 Lightning for code review and tool calling, plus an open-source model router called NeMo Switchyard.