NVIDIA has announced a 64GB version of DGX Spark, its compact system for local AI development. In its October 2 update, the company says the configuration will be available from manufacturing partners on October 23, starting at $4,999.
The new version retains the GB10 Grace Blackwell Superchip, DGX OS and NVIDIA’s AI software stack, while offering less unified memory than the 128GB configuration. Partner manufacturers named in the announcement include Acer, ASUS, Dell, Gigabyte, HP and MSI.
Two 64GB systems can be connected through their ConnectX-7 networking ports, with NVIDIA Sync Cluster Assistant helping configure the connection. NVIDIA reports up to 1.7 times the performance of one system in a specific Qwen model test; that result should not be generalized to every workload.
For developers, the choice concerns where a model runs and how much memory a project needs. Local execution can keep a supported workflow off a cloud instance, but larger models and concurrent jobs still have hardware requirements. NVIDIA also describes a model launcher arriving later in October, so it should be distinguished from software already available.
Related: CoreWeave brings NVIDIA Vera Rubin systems to its AI cloud.
Source: NVIDIA’s announcement, 2026-10-02.
