Dive Brief:
- Nvidia launched a family of models, Llama Nemotron, which are designed to provide enterprises with a foundation for AI agent development, the chipmaker said Tuesday during its annual GPU Technology Conference.
- The models were built with Meta’s Llama large language models and enhanced during post-training to improve multistep math, coding, reasoning and complex decision-making, Nvidia said. Llama Nemotron models are available as inference microservices and come in three sizes optimized for deployment variations: Nano for PCs and edge devices, Super for single GPU instances, and Ultra for multiGPU servers.
- Microsoft, SAP, ServiceNow, Accenture and Deloitte are already using or planning to use Llama Nemotron to enhance their own offerings. Accenture, for example, said its new industry-tailored agents will leverage Nvidia’s models, the professional services firm said Tuesday.
Dive Insight:
Nvidia’s latest model release expands the vendor’s agentic portfolio at a time when nearly every IT vendor is looking to do the same.
Nvidia’s AI Enterprise software platform already includes the AgentIQ toolkit, an open-source library for evaluating and simplifying development, and NIM microservices, which optimizes inference for agentic applications.
At the conference, Nvidia showcased Yum Brands’ use of NIM in its development of voice ordering AI agents.
Yum has already deployed the voice AI agents, which can understand natural language prompts, process complex menu orders and suggest add-ons, in some locations. The goal is to increase accuracy, improve customer satisfaction and reduce bottlenecks in high-volume locations. The restaurant company plans to roll the technology out at 500 locations this year.
Enterprises have high hopes for agents even as many organizations remain in the preparation phase. Business leaders have pointed to data security and privacy and legacy technology integration challenges as key adoption hurdles.
Despite the roadblocks, enterprises remain ambitious. More than 2 in 5 enterprises are planning to develop 100-plus AI agent prototypes, according to Tray.ai research published in December.
AI agent early adopters span industries. Automotive company Rolls-Royce has agents that are helping service desk workers and streamlining operations. Walmart is using an AI agent to help merchants identify supply management issues, the retailer said last month. Beauty giant Estée Lauder Companies tasked an agent with helping employees market products and analyze company archives.
Vendors are working to lower adoption barriers by releasing tools that ease development and offer customization opportunities. Enterprises are beginning to jump on board.
Over the last quarter, organizations built more than 400,000 custom agents in Microsoft’s Copilot Studio, the tech giant said earlier this month. In January, ServiceNow said it has around 1,000 customers using AI agents since introducing the technology in September.
