Nvidia debuts Llama Nemotron open reasoning models in a bid to advance agentic AI
1 min read
Summary
Nvidia is looking to compete with AI open source reasoning models with its Llama Nemotron portfolio, in response to the rising popularity of such tools in recent years.
Llama Nemotron uses Meta’s Llama models as a foundation, using algorithmic pruning to optimise compute requirements and post-training with synthetic data to enhance reasoning capabilities.
The programme offers three models, Nemotron Super, Nemotron Ultra and Nemotron Nano, which are balanced for optimal throughput and accuracy, maximum agentic accuracy and optimised for edge and smaller deployments while maintaining high reasoning accuracy, respectively.
Nvidia has also launched the Agent AI-Q blueprint, an open-source framework that allows AI agents to connect to enterprise systems and data sources, with availability scheduled for April.