Based on NVIDIA's GB10 Grace Blackwell Superchip, with a 20-core ARM-based CPU, 128GB of unified memory, a GPU with 6,144 CUDA cores, and 240W system power, the DGX Spark measures at 150x150x50.5mm and weighs just 1.2kg, making it quite an impressive little thing. It also features NVIDIA ConnectX-7 200 Gb/s networking and NVIDIA NVLink-C2C technology, providing 5x the bandwidth.

What makes the NVIDIA DGX Spark special is the NVIDIA AI software stack that comes pre-installed, giving developers access to NVIDIA AI ecosystem tools including models, libraries, and NVIDIA NIM™ microservices, enabling local workflows such as customizing Black Forest Labs’ FLUX.1 models to refine image generation, creating a vision search and summarization agent using the NVIDIA Cosmos™ Reason vision language model, or building an AI chatbot using Qwen3.

According to NVIDIA, the DGX Spark is capable of running inference on AI models with up to 200 billion parameters and fine-tune models of up to 70 billion parameters locally.

NVIDIA's CEO, Jensen Huang, made a special hand-delivery of the DGX Spark to Elon Musk at SpaceX, a nod to the first NVIDIA DGX-1 delivery to OpenAI back in 2016. NVIDIA confirmed that DGX Spark was already shipped to Anaconda, Cadence, ComfyUI, Docker, Google, Hugging Face, JetBrains, LM Studio, Meta, Microsoft, Ollama, and Roboflow.

Priced at $3,000, the DGX Spark will open up for public orders as of October 15, with several partners like Acer, ASUS, Dell, GIGABYTE, HP, Lenovo, and MSI, shipping their own versions to retailers/e-tailers.