Nvidia announced the start of shipments for DGX Spark, which it describes as the world’s smallest Artificial Intelligence supercomputer. The compact desktop system targets developers whose workloads are outgrowing conventional PCs and workstations, bringing a petaflop of Artificial Intelligence performance and 128 GB of unified memory to local labs and offices. Nvidia says DGX Spark enables inference on models with up to 200 billion parameters and local fine tuning of models up to 70 billion parameters, while also supporting the creation of Artificial Intelligence agents and advanced software stacks on site.
DGX Spark integrates the full Nvidia Artificial Intelligence platform, combining GPUs, CPUs, networking, CUDA libraries and Nvidia Artificial Intelligence software in a single system. It is accelerated by a GB10 Grace Blackwell Superchip, Nvidia ConnectX-7 200 Gb per second networking and NVLink C2C, which provides five times the bandwidth of fifth generation PCIe, along with 128 GB of CPU GPU coherent memory. The system ships with the Nvidia Artificial Intelligence software stack preinstalled, including access to models, libraries and Nvidia NIM microservices. Example local workflows highlighted by Nvidia include customizing Black Forest Labs’ Flux.1 models for image generation, creating a vision search and summarization agent using the Nvidia Cosmos Reason vision language model, and building a Qwen3 based chatbot optimized for DGX Spark.
Nvidia is positioning DGX Spark as a desktop class supercomputer that benefits from its ecosystem reach. Early recipients such as Anaconda, Cadence, ComfyUI, Docker, Google, Hugging Face, JetBrains, LM Studio, Meta, Microsoft, Ollama and Roboflow are testing, validating and optimizing their tools and models for the platform. Research groups including the NYU Global Frontier Lab previewed the system, with professor Kyunghyun Cho citing desktop access to peta scale computing that enables rapid prototyping and experimentation, including for privacy and security sensitive use cases such as healthcare.
To mark the milestone, Nvidia founder and CEO Jensen Huang hand delivered one of the first DGX Spark units to Elon Musk at SpaceX’s Starbase in Texas, drawing a connection to the 2016 delivery of the first DGX 1 to the OpenAI team. “With DGX Spark, we return to that mission, placing an Artificial Intelligence computer in the hands of every developer to ignite the next wave of breakthroughs,” Huang said. DGX Spark orders open on Oct. 15 at Nvidia.com, with partner systems available from Acer, ASUS, Dell Technologies, Gigabyte, HP, Lenovo and MSI, as well as through Micro Center in the U.S. and Nvidia channel partners worldwide.
