ServiceNow has announced the launch of Apriel Nemotron 15B, a new open-source large language model (LLM) developed in partnership with NVIDIA. Unveiled at Knowledge 2025, the model aims to redefine how enterprises use AI to manage IT, HR, and customer service operations through intelligent, real-time agents.
The Apriel Nemotron 15B model combines cost efficiency with advanced reasoning capabilities, providing a compact alternative to larger general-purpose models without compromising performance. Built with NVIDIA NeMo, NVIDIA’s Llama Nemotron open datasets, and ServiceNow’s domain-specific data, it was trained using the high-performance NVIDIA DGX Cloud on Amazon Web Services (AWS).
Compact model designed for real-time reasoning
Unlike many general-purpose LLMs that exceed a trillion parameters, Apriel Nemotron 15B uses a more focused 15 billion parameters to deliver real-time reasoning. This includes the ability to draw inferences, navigate complex rules, and support dynamic decision-making. Its smaller size allows for faster response times and lower inference costs, while still delivering enterprise-grade intelligence at scale.
Development was accelerated by post-training on NVIDIA DGX Cloud, ensuring high accuracy and fast performance across thousands of concurrent workflows. “The Apriel Nemotron 15B model — developed by two of the most advanced enterprise AI companies — features purpose-built reasoning to power the next generation of intelligent AI agents,” said Jon Sigler, executive vice president of Platform and AI at ServiceNow. “This achieves what generic models can’t, combining real-time enterprise data, workflow context and advanced reasoning to help AI agents drive real productivity.”
Closed-loop learning for smarter AI agents
As part of the launch, ServiceNow and NVIDIA also introduced a closed-loop learning architecture that allows the model to continuously improve. This setup integrates ServiceNow’s Workflow Data Fabric with NVIDIA NeMo microservices, such as NeMo Customizer and NeMo Evaluator, to fine-tune outputs using enterprise workflow data.
This approach enhances accuracy by refining AI outputs in context, with built-in guardrails to ensure data privacy and compliance. It also helps personalise agent responses over time, adapting to specific use cases and environments.
Kari Briski, vice president of generative AI software at NVIDIA, said, “Together with ServiceNow, we’ve built an efficient, enterprise-ready model to fuel a new class of intelligent AI agents that can reason to boost team productivity. By using the NVIDIA Llama Nemotron Post-Training Dataset and ServiceNow domain-specific data, Apriel Nemotron 15B delivers advanced reasoning capabilities in a smaller size, making it faster, more accurate and cost-effective to run.”
Real-world use cases and upcoming availability
During the Knowledge 2025 keynote, ServiceNow showcased a live demo featuring how the model is being used at AstraZeneca. In this example, AI agents help employees resolve workplace issues and make faster decisions, contributing to an estimated 90,000 hours in time savings for staff.
The launch follows the April debut of NVIDIA’s Llama Nemotron Ultra model, which is also based on open datasets and demonstrates strong capabilities in scientific reasoning, advanced mathematics, and programming tasks.
Apriel Nemotron 15B will power ServiceNow’s AI Agent offerings, supporting the Now LLM services and enabling broader adoption of agentic AI across industries. According to the company, businesses can expect greater digital responsiveness, improved productivity, and solutions that scale effectively with changing enterprise needs.
ServiceNow’s new model, powered by NVIDIA technology, represents a major step in bringing advanced yet efficient AI tools into real-world enterprise environments.
