NVIDIA will launch a new 64GB configuration of the DGX Spark personal AI supercomputer on October 23, giving developers a lower-memory option for running AI models locally. The system will start at $4,999 and will be available from Acer, ASUS, Dell, Gigabyte, HP and MSI.
The 64GB model retains the GB10 Grace Blackwell Superchip, DGX OS and NVIDIA’s full AI software stack found in the 128GB version. NVIDIA says it can run models with up to 100 billion parameters entirely on the device, without requiring a cloud connection.
DGX Spark combines Grace Blackwell computing, unified memory, ConnectX-7 networking and NVIDIA’s CUDA-accelerated AI software in a compact system designed for AI agents, inference, fine-tuning, data science and edge development.
Two 64GB systems can also be connected to create a 128GB cluster. NVIDIA’s Sync Cluster Assistant automatically detects connected systems, validates their configuration and sets up the ConnectX-7 network, allowing developers to scale workloads without reconfiguring their software environment.
According to NVIDIA, two systems delivered up to 1.7 times the performance of a single unit in its Qwen 3.8 27B test. The configuration can also support models with up to 200 billion parameters when the two systems are connected.
DGX Spark includes support for the NVIDIA Agent Toolkit, CUDA-X AI libraries, Nemotron open models, Ollama, vLLM and PyTorch with CUDA. Blender is also preparing a downloadable prebuilt installer for the platform.
NVIDIA Sync Model Launcher is scheduled to arrive later in October, allowing users to download and launch Qwen3.8 27B on a single DGX Spark or a connected cluster. The tool will configure the model across connected systems and set up OpenCode for use through a browser.
The NVIDIA DGX Spark 64GB will be available from Acer, ASUS, Dell, Gigabyte, HP and MSI on October 23, starting at $4,999.
