Hermès leverages advanced AI agents on NVIDIA RTX PCs and DGX Spark for self-improving innovation, showcasing cutting-edge te

Editorial illustration for Hermes deploys self‑improving AI agents using NVIDIA RTX PCs and DGX Spark

Hermes deploys self‑improving AI agents using NVIDIA RTX...

Hermes deploys self‑improving AI agents using NVIDIA RTX PCs and DGX Spark

By AI Daily Post Edited by Brian Petersen, Editor-in-Chief

May 13, 2026 • Updated: May 21, 2026 • 2 min read

Hermes is rolling out AI agents that don’t just sit idle—they stay on, take requests, plan multi‑step actions and improve themselves over time. To keep that momentum going, the company has paired the agents with NVIDIA’s DGX Spark, a compact box designed for “all‑day agentic workflows.” While the hardware is modest in size, it packs 128 GB of unified memory and a full petaflop of AI compute, enough to run 120 billion‑parameter mixture‑of‑experts models around the clock. The newer Qwen 3.6 35B model adds a leaner alternative, delivering comparable intelligence with faster throughput and room for concurrent jobs.

NVIDIA provides a “Hermes DGX Spark playbook” to streamline setup, and developers can sign up for hands‑on sessions in the “Build It Yourself” series to learn about tools like NemoClaw and OpenShell. The system is already for sale through NVIDIA’s manufacturing partners via the marketplace. In short, Hermes is betting on a hardware‑software combo that promises continuous, self‑improving AI without the need for a full‑scale data center.

NVIDIA DGX Spark is the ideal companion -- a compact, efficient standalone machine built for sustained, all-day agentic workflows.
With 128GB of unified memory and 1 petaflop of AI performance, NVIDIA DGX Spark can run 120 billion-parameter mixture-of-experts models all day. And the new Qwen 3.6 35B model delivers equivalent intelligence in a leaner footprint -- running faster and giving users the capacity to run concurrent workloads.
To maximize performance and ease of use, read the Hermes DGX Spark playbook. Plus, register for upcoming hands-on sessions in NVIDIA's "Build It Yourself" agentic AI series to learn how to build autonomous AI agents with NemoClaw and OpenShell.
NVIDIA DGX Spark is available to order from NVIDIA's manufacturing partners -- visit the marketplace.

Hermes Unlocks Self-Improving AI Agents, Powered by NVIDIA RTX PCs and DGX Spark - NVIDIA AI Blog

Why this matters We see Hermes putting self‑improving agents onto NVIDIA’s RTX PCs and the new DGX Spark. The hardware is impressive. The Spark’s 128 GB unified memory and 1 PFLOP of AI performance promise to keep agents running “all‑day” without a server farm.

For developers, that could mean fewer moving parts and lower latency when deploying continuous planning loops. Founders may like the compact, standalone form factor; it fits a rack‑mount or even a desk. Researchers, however, should note the claim that the system can handle a 120‑billion‑parameter mixture‑of‑experts model, yet the article provides no benchmark or cost analysis.

Is the performance gain sufficient to offset the expense of a DGX‑class box? The self‑improving aspect remains opaque—how agents evaluate and modify themselves is not detailed. Consequently, while the hardware appears tailored for persistent agentic workloads, the practical benefits for most AI teams are still uncertain.

We’ll watch whether Hermes’ deployment translates into measurable productivity gains or simply adds another layer of complexity.