Background
NVIDIA's RTX Spark is a line of compact desktop computers built around NVIDIA hardware for running AI locally. The first machines went on sale on 18 November 2026.[1]
What happened
NVIDIA said compact desktops from ASUS, Dell and MSI were available for purchase on 18 November 2026.[1] Mini PCs from Acer, GIGABYTE and HP, and the Microsoft Surface RTX Spark Dev Box, were said to be coming soon.[1]
Key details
The desktops use a 20-core NVIDIA Grace CPU and an NVIDIA Blackwell RTX GPU with 6,144 CUDA cores. They have up to 128GB of unified memory and deliver up to 1 petaflop of AI performance. NVIDIA said they can run 120-billion-parameter large language models locally.[1]
The ASUS ProArt GR1X mini PC is built to run personal AI agents 24/7.[1]
Developers can set up agents with Hermes Agent or NVIDIA NemoClaw blueprints, and they can fine-tune supported models with Unsloth.[1] NVIDIA Switchyard, now available in Hermes, automatically routes agent tasks between local and cloud models.[1]
Adobe Premiere, Photoshop, Illustrator and InDesign gained native Windows on Arm versions or RTX acceleration for RTX Spark.[1]
Impact
The launch combines local model hosting on the desktop with routing that can send agent tasks to cloud models when needed.[1]
Terms used in this entry
Unified memory means the CPU and GPU share one pool of memory, so data does not have to be copied between two separate memories. A CUDA core is one of the parallel processing units inside an NVIDIA GPU, and the number of cores is one rough indicator of how much parallel work the GPU can do. A petaflop is one thousand trillion floating-point operations per second, a common unit for AI hardware performance.
A large language model with 120 billion parameters has 120 billion adjustable numbers that together determine how it responds. Running such a model locally means the computation happens on the user's own machine rather than in a remote data center. Fine-tuning means further training an existing model on a smaller, specific set of examples so that it performs better on a particular kind of task.
Windows on Arm is a version of Windows built for processors that use the Arm instruction set, rather than the x86 processors used in most Windows PCs. Native versions of an application run directly on that processor. RTX acceleration means an application uses the GPU's features to speed up work such as image processing.
Sources
- 1NVIDIA RTX Spark Compact Desktops Out NowNVIDIA Blog
Cite this page: NVIDIA RTX Spark compact desktops go on sale. TellWorld, https://www.tellworld.net/news/rtx-spark-desktops-2026. Updated 2026-11-20