Extra Nvidia information from SIGGRAPH
The RTX Professional server wasn’t the one information on the present. Nvidia additionally introduced two new fashions of its Nemotron mannequin household – the Nemotron Nano 2 and Llama Nemotron Tremendous 1.5 – with superior reasoning capabilities for constructing smarter AI brokers.
Nemotron is a household of enterprise-ready, open giant language fashions (LLM) designed to reinforce agentic AI for duties requiring subtle reasoning, instruction following, coding, software use, and multimodal (textual content+imaginative and prescient) understanding.
These fashions ship excessive accuracy for his or her relative measurement in areas equivalent to scientific reasoning, coding, software use, instruction following and chat, in keeping with Nvidia. They’re designed to imbue AI brokers with deeper cognitive talents and assist AI programs discover choices, weigh choices and ship outcomes inside outlined constraints.
Nvidia claims Nemotron Nano 2 achieves as much as six instances greater token era throughput in comparison with different fashions its measurement. Llama Nemotron Tremendous 1.5 provides top-tier efficiency and leads in reasoning accuracy, making it appropriate for dealing with complicated enterprise duties.
Additionally, Nvidia is empowering robotics and machines to “see” and react to what they see with new AI fashions that may ingest visible data and take into consideration mentioned data.
The seller simply introduced Cosmos Purpose, a brand new open, customizable 7 billion-parameter reasoning Imaginative and prescient Language Fashions, or VLMs. VLMs permit robots and imaginative and prescient brokers to consider what they see, similar to a human. So far, robots have had the flexibility to “see,” however their response to what they noticed was extraordinarily restricted. A VLM gives robotics with the flexibility to consider their actions.
