According to Dongcha Beating monitoring, NVIDIA is doubling down on Nemotron 4, aiming to make it one of the world's most powerful open models. The Information cited multiple project employees, stating that the largest model in the series is expected to have at least 1 trillion parameters, about twice the size of Nemotron 3 Ultra. The specific specifications and release date are yet to be determined, and final training has not yet begun.
NVIDIA hopes to drive GPU demand with more powerful open models. The company has significantly increased its investment in self-developed models' computing power in recent years. As of April this year, the commitment to cloud service procurement for the next few years has increased to $28 billion, about three times the amount from a year ago, although this also includes other R&D purposes.
Nemotron 4 has also brought in a group of external companies for joint development. Cursor, Mistral, Thinking Machines, Reflection AI, among others, will provide training data, evaluation, and model design. Prime Intellect has even contributed 300,000 simulated environments.
The issue is that many of these companies are already NVIDIA's customers or invested enterprises. Leading model companies like OpenAI are also among its largest GPU buyers. Now NVIDIA is venturing into top-tier open models itself, aiming to both expand the open model market and compete with its own customers and partners.
