Nvidia releases new open-source AI model, claiming "faster, cheaper and smarter"

NVIDIA (NVDA. O) on Monday unveiled a slew of new open-source artificial intelligence models that it says will be faster, cheaper and smarter than its predecessors. NVIDIA is mostly known for supplying chips that companies such as OpenAI use to train and profit from its closed-source models. But NVIDIA also offers a large number of its own models, covering everything from physics simulations to autonomous cars, that are made available in the form of open-source software for researchers or other companies, such as Palantir Technologies, that have integrated NVIDIA's models into their products. NVIDIA on Monday unveiled its third-generation "Nemotron" large language model, which is mainly aimed at tasks such as writing and programming. The smallest of these models, the Nemotron 3 Nano, was released on the same day, and two more larger versions will be launched in the first half of 2026. Meanwhile, there are reports that Meta Platforms (META. O) is considering a move to closed-source models, making Nvidia one of the leading providers of open-source models in the United States.