ServiceNow and Nvidia have introduced Apriel 2.0, the latest version of their jointly-developed Apriel Nemotron open model. It boasts improved reasoning, native support for multimodal input, and built-in data and safety guardrails. However, unlike its predecessor, which was a general-purpose foundation model, Apriel 2.0 is explicitly designed for autonomous and semi-autonomous agents operating in regulated industries. Moreover, this next-generation AI is post-trained on both Nvidia and ServiceNow data.
“The next wave of AI is about more than innovation. It’s about execution—how fast and how responsibly enterprises can put advanced intelligence to work,” Pat Casey, ServiceNow’s chief technology officer and executive vice president of DevOps, remarks in a release. “Our collaboration with Nvidia is built around that idea. By releasing open models with best-in-class reasoning, we can deliver AI that’s efficient, trusted, and built to scale.”
Subscribe to The AI Economy
The Apriel Evolution
Earlier this spring, ServiceNow introduced the first Apriel models with Apriel-5B-Base and Apriel-5B-Instruct. These compact language models were designed for general-purpose tasks—answering questions, retrieving information, generating content, assisting with code, reasoning, and creative writing. They were also trained on ServiceNow’s own data.
In May, ServiceNow partnered with Nvidia to release Apriel Nemotron 15B, a larger language model focused on advanced reasoning, powered primarily by Nvidia’s data. At the time, ServiceNow’s Executive Vice President of Platform and AI, Jon Sigler, shared that Apriel Nemotron 15B would power intelligent AI agents capable of making context-aware decisions, adapting to complex workflows, and delivering personalized outcomes at scale.
And just this month, ServiceNow announced Apriel-1.5-15B-Thinker, an open, multimodal reasoning model trained with a data-centric mid-training recipe and capable of fitting on a single GPU. This was the first model to benefit from Nvidia’s Nemotron open model collection.
Where Apriel 2.0 Fits In
Apriel 2.0 represents a convergence of everything ServiceNow and Nvidia have been building toward—a model that combines the scale and reasoning strength of Apriel Nemotron with the compact efficiency of the original Apriel. Despite its power, it’s optimized to run on a single GPU, and unlike its predecessor, Apriel 2.0 isn’t a general-purpose foundation model; it’s purpose-built for reasoning within regulated industries such as financial services, healthcare, and telecommunications.
“Most AI models act like black boxes; Apriel is built differently,” Joe Davis, ServiceNow’s executive vice president of platform engineering and the AI technology group, tells me in an email. “It’s a multimodal reasoning model that can ‘show its work,’ transparently mapping how it reaches conclusions across text, documents, and data. That explainability is essential for auditability and trust in federal and highly regulated environments.”
The company confirmed the model has 15 billion parameters but hasn’t shared benchmark results. It did, however, highlight Apriel 2.0’s ability to interpret screenshots, forms, and diagrams, giving it a fuller picture of enterprise workflows.
With its low latency and multi-step reasoning, Apriel 2.0 is designed to power autonomous agents capable of resolving complex service requests end to end, from diagnosing IT outages to generating and deploying fixes.
“[Apriel 2.0] operates entirely on customers’ governed data, not public datasets, with policy and reasoning guardrails applied before any output is produced. Running on ServiceNow’s authorized infrastructure, it inherits each organization’s existing permissions, governance rules, and audit trails, ensuring AI outputs stay within established compliance frameworks,” Davis added.
ServiceNow plans to launch Apriel 2.0 in Q1 2026.
Expanding Nvidia Partnership
ServiceNow’s announcement is timed to Nvidia’s GTC conference and is meant to demonstrate how the two companies are furthering their long-standing partnership. But beyond the unveiling of Apriel 2.0, ServiceNow is also revealing that it’s integrating into Nvidia’s AI Factory and reference architectures.
By tying into Nvidia’s AI Factory blueprint, ServiceNow can embed its AI platform natively into the environments where enterprises generate and refine their own models. For Nvidia, this collaboration gives its industrial-scale AI infrastructure a practical enterprise outlet—ServiceNow’s customers span regulated and high-compliance industries and are the same sectors Nvidia wants to appeal to with its AI Factory model.
The integration is expected to also launch in Q1 2026.
Featured Image: ServiceNow CEO Bill McDermott (second from left) chats with Nvidia CEO Jensen Huang (far left) and “Acquired” podcast hosts Ben Gilbert and David Rosenthal at the Knowledge 2025 conference in Las Vegas on May 6, 2025. Photo credit: Ken Yeung
Subscribe to “The AI Economy”
Exploring AI’s impact on business, work, society, and technology.
Leave a Reply
You must be logged in to post a comment.