NVIDIA DGX Spark 64GB Launches at US$4,999, Built to Be Clustered

NVIDIA Halves the DGX Spark’s Memory for a US$4,999 Entry Point and Pitches Clustering as the Way Up

NVIDIA has announced a 64GB configuration of the DGX Spark, its compact desktop AI system, starting at US$4,999. The new model is the same machine as the existing DGX Spark with half the unified memory: it keeps the GB10 Grace Blackwell Superchip, DGX OS, the NVIDIA AI software stack and the built-in NVIDIA ConnectX-7 networking. It goes on sale Friday, October 23, exclusively through Acer, ASUS, Dell, GIGABYTE, HP and MSI. NVIDIA will not sell a Founders Edition of the 64GB model, so the NVIDIA-branded unit remains a 128GB-only product.

The launch arrives in the middle of the memory crunch. NVIDIA also told that they will be raising the price of the 128GB DGX Spark to US$6,950. That model launched in October 2025 at US$3,999 and moved to US$4,699 after a price increase in February 2026. At US$4,999, the new 64GB entry model costs 25% more than the 128GB unit did at launch.

DGX Spark 64GB vs DGX Spark 128GB

NVIDIA has not published an SSD capacity for the 64GB model. As with every DGX Spark, partners price their systems around the SSD they ship, so US$4,999 is the base figure and higher-storage configurations will cost more.

For our measured performance and power results on the 128GB hardware, plus a simulated 64GB memory limit, read our GIGABYTE AI TOP ATOM 128GB DGX Spark review.

What a Single 64GB Unit Runs

NVIDIA quotes support for models of up to 80 billion to 100 billion parameters on a single DGX Spark 64GB, running entirely on device. The practical ceiling depends on the model and how heavily it is quantized. NVIDIA positions the system for agents, inference, fine-tuning, data science and edge development, with the following supported out of the box:

• NVIDIA tooling: NVIDIA Agent Toolkit, CUDA-X AI libraries and Nemotron open models.
• Runtimes: Ollama, vLLM and PyTorch with CUDA, with llama.cpp and LM Studio also listed as supported inference frameworks.
• Creator apps: Blender is among the first major creator applications to support the platform, with a prebuilt, downloadable installer coming soon.

Stack ’em High with Clusters

Alongside the 64GB model, NVIDIA announced NVIDIA Sync Cluster Assistant, a feature of the NVIDIA Sync app that links DGX Spark systems into a cluster. Two units connect directly with a QSFP cable through their ConnectX-7 ports over the 200 GbE fabric. According to NVIDIA, a pair of 64GB units pools memory to 128GB, expands model support to up to 200 billion parameters, and delivers twice the memory bandwidth. In NVIDIA’s Qwen 3.8 27B test, two clustered 64GB systems delivered up to 1.7x the performance of a single system.

• Detection: Cluster Assistant finds connected DGX Spark units.
• Validation: It checks each device’s configuration before joining it to the cluster.
• Networking: It configures the ConnectX-7 network. Every node runs the same NVIDIA software stack, so nothing needs reconfiguring when moving from one unit to two.

NVIDIA’s scaling chart carries the idea past two units:

At the end of October, NVIDIA will add NVIDIA Sync Model Launcher. It downloads and launches Qwen3.8 27B on a single DGX Spark or a cluster, configures the model to run across connected devices, and makes it accessible from the user’s laptop. The launcher also sets up OpenCode to use the model, so developers can start coding in their browser.

The Workflows NVIDIA Is Targeting

• Always-on agents: Keep a coding or research agent running on DGX Spark to review code, analyze documents or carry out multistep tasks. A cluster adds capacity for larger models, longer context windows or several agents at once.
• Offloading AI from an everyday PC: Run a language or image-generation model on DGX Spark while the agent or creative application runs on a laptop or desktop, leaving the PC free for other work.
• Scaling up: When a task outgrows one unit, two DGX Spark 64GB systems connected through Cluster Assistant pool their memory to 128GB, and the same workflow carries over without changes to the software environment.

Getting Started

NVIDIA’s setup path is to install a supported inference framework (llama.cpp, Ollama, vLLM or LM Studio), download the recommended local model for the workflow, and, for two units, connect them through their ConnectX-7 ports and launch Cluster Assistant. Playbooks are available at build.nvidia.com/spark, including the NemoClaw, OpenClaw, Hermes Agent and OpenShell agentic playbooks. Three playbooks are coming soon to 64GB devices:

• Serve LLMs With vLLM
• Run OpenClaw With a Local LLM
• Connect Multiple DGX Sparks for Distributed Workloads

Full product details are on NVIDIA’s DGX Spark product page.

The RTX Spark Connection

The DGX Spark 64GB lands days before Microsoft’s October 7 event in San Francisco, scheduled for 10:00 AM PT (1:00 AM on October 8, Philippine time) and focused on local AI on the PC. Microsoft CEO Satya Nadella, Windows and Surface chief Pavan Davuluri and NVIDIA CEO Jensen Huang are confirmed to attend, and NVIDIA RTX Spark is expected to get its formal launch there. NVIDIA’s own announcement states that new Windows PCs powered by RTX Spark are coming this month from Acer, ASUS, Dell, HP, Lenovo, Microsoft and MSI. We expect the DGX Spark 64GB to be part of that story, since NVIDIA pitches it as the box that runs the models while an everyday PC runs the apps.

Our Take

NVIDIA is pushing clustering hard with this launch. The 64GB model gets developers in at a lower price, and pooling more units is the upgrade path NVIDIA is selling, which reads to us as Jensen Huang’s “the more you buy, the more you save” brought to the desk. The math cuts both ways. Two 64GB units at US$4,999 each come to US$9,998, above the reported US$6,950 for a single 128GB unit, for the same 128GB of memory. In exchange, NVIDIA’s chart shows the pair delivering twice the FP4 compute and memory bandwidth of one unit.

The other reason for this SKU is the RAM crisis. Halving the memory lets NVIDIA and its partners keep the DGX Spark line moving at an entry price buyers can still reach, while memory and SSD costs continue to push the 128GB model up.

Philippine Pricing and Availability

The DGX Spark reached the Philippines later than other markets, and local pricing has moved with foreign exchange rates as well as RAM and SSD costs. We have seen partners quote the DGX Spark 128GB at up to PHP690,000, and those quotes may change with the reported price increase on the 128GB model. Rollout of the 64GB model will vary by region, and local pricing and availability have not been announced. We will update this story once partners confirm Philippine details.