March 20 news.NvidiaNew Release Llama Nemotron Series Open Reasoning AI ModelsThe "Intelligent Body AI" system is supported.It can independently reason, plan, and execute multi-step tasks for robotics, automation, and decision optimization scenarios, significantly reducing reasoning costs and helping enterprises achieve AI autonomy upgrades.
The Llama Nemotron series of open inference AI models are available in Nano, Super, and Ultra scales, which 1AI cites as briefly described in the blog post below:
- Nano: Designed for PCs and edge devices to provide highly accurate inference.
- Super: Single GPU deployment for top accuracy and processing speed.
- Ultra: Multi-GPU server optimized to ensure the highest accuracy for complex tasks.
With three different scales for different deployment environments, deep inference modes can be activated on-demand, saving arithmetic and reducing inference costs as they simplify tasks, and the Llama Nemotron family of open inference AI models also supports a full range of scenarios, from datacenter to edge computing, through the NVIDIA NIM microservices.
NVIDIA also introduced a new member of Cosmos Nemotron, a vision-focused language model (VLM) that parses image, video, and sensor data to extend the capabilities of NVIDIA's previous Cosmos World Foundation Model (WFM) platform, which is used in autonomous driving, industrial robotics, and other applications that require multimodal reasoning. It enables collaborative "text + vision" decision-making.
