Nvidia Introduces More Affordable DGX Spark Model with Reduced Memory

Memory Choice Affects Workload Suitability

Nvidia has unveiled a new variant of its DGX Spark AI workstation featuring 64 GB of unified memory, priced at $4,999. This version launched alongside the existing 128 GB model, which carries a $1,000 higher price tag. The announcement was reported by Michael Kan of PCMag and shared via Techmeme on October 2, 2026.

The DGX Spark is designed as a compact, desktop-sized system for AI development and inference tasks. Both configurations use the same underlying architecture, differing only in memory capacity. The 64 GB version targets developers and small teams who require substantial AI computing power but do not need the full memory footprint of the higher-end model. By offering a lower-cost entry point, Nvidia aims to broaden access to its AI hardware platform.

Who Is This Model Intended For?

The 64 GB configuration is best suited for training smaller language models, running inference on moderate-sized AI applications, or handling data preprocessing tasks. Users working with large-scale models or extensive datasets may still need the 128 GB version to avoid memory bottlenecks. Nvidia positions the Spark series as a bridge between consumer GPUs and its larger DGX systems, providing a balance of performance and affordability for professional AI workflows.

This tiered pricing strategy allows Nvidia to serve a wider audience, from individual researchers to startup AI labs. The $4,999 price makes the DGX Spark more accessible compared to enterprise-grade systems that often cost tens of thousands of dollars. It reflects Nvidia’s effort to democratize AI development tools while maintaining performance standards. The unified memory architecture ensures efficient data sharing between CPU and GPU components, reducing latency in AI pipelines.

What is the main difference between the two DGX Spark models? The only difference is the amount of unified memory: 64 GB in the lower-cost model and 128 GB in the higher-priced version, with all other specifications remaining identical.

Frequently Asked Questions

Why would someone choose the 64 GB model over the 128 GB version? The 64 GB model is sufficient for many AI development tasks, such as prototyping models or running inference, and offers a $1,000 savings for users who do not require extensive memory capacity.

Does the lower memory version affect overall system performance? Performance remains consistent for workloads that fit within 64 GB of memory; however, larger models or datasets may exceed this limit and require the 128 GB configuration to operate efficiently.