ServiceNow and Nvidia's new reasoning AI model raises the bar for enterprise AI agents

Elyse Betters Picaro / ZDNET

Many person dubbed this twelvemonth "the twelvemonth of AI agents," arsenic these AI systems that tin transportation retired tasks for users are particularly utile for optimizing endeavor workflows. At ServiceNow's yearly Knowledge 2025 conference, the institution unveiled a caller exemplary successful concern with Nvidia to beforehand AI agents.

Apriel Nemotron 15B

On Tuesday, ServiceNow and Nvidia launched Apriel Nemotron 15B, a new, open-source reasoning connection exemplary (LLM) built to present little latency, little inference costs, and agentic AI. According to the release, the exemplary was trained connected Nvidia Nemo, the Nvidia Llama Nemotron Post-Training Dataset, and ServiceNow's domain-specific data.

Also: Nvidia's 70+ projects astatine ICLR amusement however earthy spot powerfulness is cardinal to AI's acceleration

The biggest takeaway of the exemplary is that it packages precocious reasoning capabilities successful a smaller size. This makes the exemplary cheaper and faster to tally connected Nvidia GPU infrastructure arsenic an Nvidia NIM microservice portion inactive delivering the enterprise-grade quality companies are looking for.

The institution shares that Apriel Nemotron 15B shows promising results for its exemplary class successful benchmark testing, confirming that the exemplary could beryllium a bully acceptable for supporting agentic AI workflows.

Also: Will synthetic information derail generative AI's momentum oregon beryllium the breakthrough we need?

Reasoning capabilities are particularly important erstwhile utilizing agentic AI because, successful these automated experiences, AI performs tasks for the extremity idiosyncratic successful assorted settings. Since it is performing tasks without quality direction, it needs to bash immoderate processing oregon reasoning of its ain to find however to proceed best.

Joint information flywheel architecture 

In summation to the model, the 2 companies besides unveiled a associated information flywheel architecture -- a feedback loop that collects information from interactions to further refine AI models. The architecture integrates ServiceNow Workflow Data Fabric and prime Nvidia NeMo microservices, according to the release.

Also: Nvidia launches NeMo bundle tools to assistance enterprises physique customized AI agents

This associated architecture allows companies to usage endeavor workflow information to further refine their reasoning models portion besides having the indispensable guardrails successful spot to support customers, guarantee the information is processed successful a unafraid and timely manner, and springiness them the power they want. Ideally, this would provender into the instauration of highly personalized, context-aware AI agents, according to the company.

Get the morning's apical stories successful your inbox each time with our Tech Today newsletter.