Bringing the Spark to Europe: Local Intelligence Powers Next Wave of Agents

NVIDIA introduced RTX Spark, a new AI-powered hardware platform for Windows PCs designed to run personal agents, accelerate content creation, and support high-performance gaming with up to 128GB of unified memory and 1 petaflop of AI compute. The platform enables fully local, private AI operations through the Underdog desktop agent app, which uses the Husky model-specific inference engine to ensure user data never leaves the device. RTX Spark laptops from ASUS, Dell Technologies, HP, Lenovo, Microsoft, and MSI are available now, with Acer models expected soon.

DGX Spark, launching October 23, targets developers and enthusiasts with 64GB systems capable of running models like Nemotron 3.5 Bolt 74B 24GB and Qwen3.8 27B. The platform supports clustering up to eight units via NVIDIA ConnectX-7 networking, enabling distributed development and concurrent agent sessions with 1TB of unified memory. New software updates include the NVIDIA Sync Model Launcher for simplified local AI access and a Wi-Fi Assistant for network management in appliance mode.

DGX Station delivers enterprise-grade AI performance with 748GB of coherent memory and up to 20 petaflops of AI compute, supporting advanced models like GLM 5.3 and serving over 30 users. The system integrates with enterprise management tools such as Splunk, Microsoft Intune, and Tanium to streamline deployment and reduce risk. DGX Station for Windows, expected in December, will extend local AI into Windows-native applications with support for professional tools from companies like Dassault Systèmes and Siemens.

NVIDIA also previewed DSX OS running on DGX Station to simulate AI factory operations and demonstrated a local token factory supporting over 30 developers. The company released open-source reference workflows for refining GR00T 1.7 models and training robotics applications. DGX Station systems are shipping now from multiple vendors, with Windows-based systems expected later this year.