Nvidia Developing Nemotron 4 AI Models With at Least 1 Trillion Parameters to Challenge Top Open-Source Rivals
Summary
- Nvidia is developing Nemotron 4 with at least 1 trillion parameters to compete with leading open-source models.
- Nvidia has not set a release date for Nemotron 4, but employees said the model could be ready as early as late fall.
- Nvidia said it formed a coalition with Microsoft and others for AI safety and cybersecurity tools and unveiled Nemotron 3.5 Lightning and NeMo Switchyard.
Forecast Trend Report by Period
Report says largest Nemotron 4 model will have at least 1 trillion parameters
Rare US open-source model draws attention
Nvidia is developing a new family of AI models called Nemotron 4 aimed at competing with the world’s leading open-source models.
The largest Nemotron 4 model under development will have at least 1 trillion parameters, according to a report by The Information cited by Reuters on Aug. 11.
Nvidia has not set a release date for Nemotron 4 and has not completed final training. Employees told the outlet the model could be ready as early as late fall.
Open-source models have drawn increased attention this year after China’s Kimi K3 came close to the performance of top AI models from Anthropic and OpenAI. Most US AI labs, however, still do not release open-source models.
Interest has also risen after recent hacking cases involving autonomous AI agents were disclosed, because open-source models face no restrictions on cybersecurity use.
Last month, Nvidia formed a coalition with Microsoft and other technology companies to develop and share AI safety and cybersecurity tools. The group also issued a letter supporting open-weight models to help prevent innovation from moving overseas.
Separately, Nvidia on Aug. 11 unveiled Nemotron 3.5 Lightning, a new product for code review, tool use, security alert monitoring, answering billing-related questions and other tasks. It also released NeMo Switchyard, an open-source model-routing library designed to automatically assign AI tasks to the most suitable model.
Open-source models disclose the full development process, including source code, training data, data preprocessing methods, training scripts and model architecture. Open-weight models, by contrast, do not disclose the full training process or data, but do release the weights, or parameters, of trained models so anyone can run or fine-tune them.
Kim Jung-a, contributing reporter, Hankyung.com, kja@hankyung.com
Korea Economic Daily
hankyung@bloomingbit.ioThe Korea Economic Daily Global is a digital media where latest news on Korean companies, industries, and financial markets.