NVIDIA is expanding its DGX Spark system lineup with a new 64GB unified memory configuration, available from 23 October through partner manufacturers Acer, ASUS, Dell, Gigabyte, HP, and MSI. The 64GB variant retains the same GB10 Grace Blackwell Superchip, DGX OS, and full NVIDIA AI software stack found in the 128GB model, offering developers and researchers a more accessible entry point into local AI computing without sacrificing core capabilities.
NVIDIA DGX Spark 64GB
With 64GB of memory at its disposal, this new DGX Spark configuration supports models of up to 100 billion parameters running entirely on device, enabling private, cloud-free inference, fine-tuning, data science, and edge development.
It comes pre-loaded with NVIDIA Agent Toolkit, CUDA-X AI libraries, Nemotron open models, and popular runtimes including Ollama, vLLM, PyTorch with CUDA, and LM Studio, allowing developers to go from power-on to running models within minutes. Blender is among the first major creative applications confirmed to support the platform, with a prebuilt downloadable installer arriving soon.
The system includes a built-in NVIDIA ConnectX-7 NIC, which serves as the foundation for its multi-node clustering capability: two DGX Spark 64GB units can be connected directly via a QSFP cable, pooling their memory to 128GB and expanding model support to up to 200 billion parameters. The configuration also delivers twice the memory bandwidth and up to 1.7x the performance compared to a single unit, based on NVIDIA’s internal testing.
The NVIDIA Sync Cluster Assistant handles the networking setup automatically. Once two units are connected, the tool detects the devices, validates their configuration, and sets up the ConnectX-7 network without requiring manual intervention. Both nodes run the same NVIDIA software stack, so no reconfiguration is needed when scaling from one unit to two.
Also arriving at the end of October, the NVIDIA Sync Model Launcher will allow users to download and deploy models such as Qwen3.8 27B on a single system or a cluster through a simplified graphical interface. The launcher will also configure OpenCode to use the local model, enabling browser-based coding workflows.
The NVIDIA DGX Spark 64GB is available starting 23 October from Acer, ASUS, Dell, Gigabyte, HP, and MSI, with pricing starting at USD$4,999. This configuration is exclusive to manufacturer partners and is not available directly through NVIDIA.
Pokdepinion: Nearly five grand to run AI locally, but hey, at least your data stays on your desk and not in someone else’s server farm.
