A new starting point for local AI
NVIDIA announced a new 64GB unified memory version of its DGX Spark personal AI supercomputer. The device will be available this month through manufacturing partners including Acer, ASUS, Dell, Gigabyte, HP and MSI. Keeping the same GB10 Grace Blackwell chip, DGX OS and NVIDIA AI software stack as the existing 128GB model, the new version is positioned at a more accessible price point.
What runs on the device?
The 64GB configuration can run models with up to 100 billion parameters and the agentic applications built on them entirely on device. This lets developers experiment with models and their own data without turning to a cloud instance for every task. The device supports NVIDIA Agent Toolkit, CUDA-X AI libraries, Nemotron open models and popular runtimes like Ollama, vLLM and PyTorch out of the box.
What happens when two units are combined?
Every DGX Spark includes a built-in NVIDIA ConnectX-7 network card. Two units can connect directly with a QSFP cable, pooling their memory to 128GB and expanding model support to up to 200 billion parameters. In NVIDIA's Qwen 3.8 27B test, two clustered 64GB systems delivered up to 1.7x the performance of a single system. The Cluster Assistant feature in the NVIDIA Sync app detects connected units and configures them automatically; no changes to the software stack are needed when scaling.
