NVIDIA will launch a new 64GB version of its DGX Spark personal AI system this October.
The new configuration will be available from Acer, ASUS, Dell, Gigabyte, HP, and MSI. It uses the same GB10 Grace Blackwell Superchip, DGX OS, and NVIDIA AI software stack as the 128GB version.
The 64GB DGX Spark can run models with up to 100 billion parameters on-device, allowing developers to work with AI models without relying entirely on cloud services.
NVIDIA also said two 64GB DGX Spark units can be connected to create a 128GB system. Through NVIDIA Sync Cluster Assistant, the two devices can share memory and workloads without requiring developers to manually reconfigure the software environment.
In NVIDIA’s test using Qwen3.8 27B, two connected 64GB systems delivered up to 1.7 times the performance of a single unit.
The system includes NVIDIA ConnectX-7 networking and supports AI development tools and frameworks such as PyTorch, CUDA-X AI, Ollama, vLLM, and llama.cpp. NVIDIA also plans to support applications such as Blender.
NVIDIA Sync will also receive a Model Launcher feature later this month. It will allow users to download and run supported models, including Qwen3.8 27B, on a single DGX Spark or a connected two-unit setup.
The company lists several use cases for the 64GB model, including running coding and research agents locally, hosting AI models for use by other PCs, and adding a second unit when a workload requires more memory or computing resources.
NVIDIA DGX Spark 64GB Availability
The NVIDIA DGX Spark 64GB will be available starting October 23 through Acer, ASUS, Dell, Gigabyte, HP, and MSI.