NVIDIA Unveils RTX Spark, DGX Spark & Station Updates at GTC Berlin

Disclosure: this post contains affiliate links. If you buy through them we may earn a commission at no extra cost to you. Learn more.

TL;DR: At GTC Berlin, NVIDIA is showcasing its full lineup of local AI hardware including RTX Spark for personal agents on Windows PCs, DGX Spark for developers on Linux, and DGX Station for enterprise teams. Key announcements include a new DGX Spark 64GB variant launching October 23, DGX Station for Windows arriving in December, and support for new AI models including Recursive Flint coding model and Nemotron Bolt E4B 4B.

What's New

NVIDIA is presenting its complete local AI hardware portfolio at GTC Berlin on October 21, 2026, with significant updates across three product lines designed for different use cases.

RTX Spark launched last week on October 16 and brings personal agents to Windows PCs with up to 128GB of unified memory and 1 petaflop of AI compute. The platform supports accelerated content creation and 1440p 100 frames-per-second gaming. NVIDIA founder and CEO Jensen Huang demonstrated how AI agents on RTX Spark can train Hugging Face's Microduck robot using NVIDIA Isaac Lab, Hermes Agent, and Nemotron 3.5 Bolt 74B. The Underdog desktop agent app, powered by its Husky model-specific inference engine, is now available for RTX Spark, running models entirely on-device for private AI without data leaving the PC. Many RTX Spark laptops feature an unlocked bootloader for developer customization.

DGX Spark receives a new 64GB variant launching Friday, October 23, capable of running models including Nemotron 3.5 Bolt 74B 24GB, Qwen3.8 27B, and Gemma 4 26B-A4B. The platform now supports clusters of up to eight units through built-in NVIDIA ConnectX-7 networking and NVIDIA Sync Cluster Assistant, delivering 1TB of unified memory for frontier-class models like GLM 5.3. The October software update adds Wi-Fi Assistant for network switching in appliance mode and DGX Spark Skills for workflow guidance. NVIDIA Sync Model Launcher arrives later this month, simplifying model downloads and container deployment.

DGX Station offers 748GB of coherent memory and up to 20 petaflops of AI performance, serving 30+ users at cloud-comparable speeds. New integrations with Absolute Security, Splunk by Cisco, Fleet Device Management, Microsoft Intune, Perforce Puppet, Progress Software Chef, and Tanium enable enterprise IT management. DGX Station for Windows launches in December with support for manufacturing applications from nTop, Dassault Systèmes, Keysight Technologies, MathWorks, Rockwell Automation, Siemens, Synopsys, Unity, and Altera. The platform will support agent harnesses including Hermes Agent, Perplexity, and Aible.

New Models and Software: Recursive Superintelligence released Recursive Flint, a 40-billion-parameter mixture-of-experts coding model with 2.5 billion active parameters, available under Apache 2.0 license. NVIDIA added day-zero support for Nemotron Bolt E4B 4B (launching October 23), GLM, DeepSeek, Qwen, MiniMax, Liquid AI's LFM2.5-24B-A2B, and FLUX.3 Klein diffusion model. NVIDIA Cloud Functions is coming in early access to Linux-based DGX Station. Three new open source reference workflows for robotics bring GPT-6 Astra, NVIDIA Isaac Lab, Isaac Lab-Arena, and Cosmos to DGX Station.

Pricing and Availability

Pricing information was not disclosed. RTX Spark laptops are available now from ASUS, Dell Technologies, HP, Lenovo, Microsoft, and MSI, with Acer coming soon. DGX Spark 64GB systems begin rolling out from partners on Friday, October 23, 2026. DGX Spark systems are available from Acer, ASUS, Dell Technologies, GIGABYTE, HP, Lenovo, MSI, and NVIDIA.

DGX Station systems are shipping today from ASUS, Dell Technologies, GIGABYTE, HP, MSI, Supermicro, and Exxact. DGX Station for Windows systems are expected to begin shipping in December 2026 from ASUS, Dell Technologies, GIGABYTE, HP, MSI, and Supermicro.

How to Try It

Developers can download Underdog for RTX Spark now. RTX Spark playbooks and the DGX Spark cluster playbook are available to help users get started. Recursive Flint model weights can be downloaded on Hugging Face. Developers can post-train models using NeMo AutoModel for supervised fine-tuning and LoRa, as well as NeMo RL for optimization. NVIDIA provides local AI inference recipes to test the latest optimizations.

Who It's For

RTX Spark targets individual creators, developers, and enthusiasts who need personal AI agents running locally on Windows PCs. DGX Spark is designed for developers building and scaling applications on Linux, particularly small teams sharing clusters as backends. DGX Station serves enterprise teams running frontier workloads, robotics teams needing dedicated compute for post-training and simulation, and AI factory development teams requiring local iteration capabilities.

Alternatives

For teams seeking local AI infrastructure alternatives, organizations might consider cloud-based development platforms or other enterprise GPU solutions. Replit offers cloud-based development environments, though without the local privacy benefits. Workflow automation platforms like Make can complement these AI development tools for enterprise teams building agent-based workflows.