NVIDIA Introduces New 64GB DGX Spark for Local Artificial Intelligence
NVIDIA has expanded its personal AI computing lineup with a new 64GB version of the DGX Spark, giving developers, researchers and AI enthusiasts another option for running advanced artificial intelligence models directly on a local machine.
The new system starts at $4,999 and is scheduled to become available from major hardware partners on October 23, 2026.
What Is NVIDIA DGX Spark?
DGX Spark is a compact AI computer designed specifically for artificial intelligence development and experimentation.
Unlike a conventional desktop PC, the system combines high-performance AI computing hardware with NVIDIA's software ecosystem, allowing developers to run AI models, inference workloads, fine-tuning tasks and AI agents locally.
The new 64GB configuration retains NVIDIA's GB10 Grace Blackwell Superchip, along with the company's AI software stack and DGX operating system.
Supports AI Models With Up to 100 Billion Parameters
According to NVIDIA, the 64GB DGX Spark can run models containing up to approximately 100 billion parameters entirely on the device.
This capability is aimed at developers who want to experiment with increasingly sophisticated AI models without sending their data to a remote cloud service for every task.
Local processing can be particularly useful when developers are working with private datasets, proprietary software or applications where controlling data location is important.
Designed for Local AI Agents
One of the main use cases highlighted by NVIDIA is the development and operation of AI agents.
AI agents can perform multi-step tasks such as analyzing documents, reviewing software code, conducting research and interacting with applications.
Running these workloads locally can provide developers with greater control over the computing environment and data used by the agent.
NVIDIA says DGX Spark can run local agent workflows without requiring constant cloud connectivity.
Same Core Platform as the Larger Model
The new 64GB configuration uses the same core GB10 Grace Blackwell platform found in the larger DGX Spark system.
It also includes NVIDIA's ConnectX-7 networking technology, DGX OS and the company's AI software ecosystem.
This means the new system is designed to provide a similar development environment while offering a lower-memory configuration.
Two DGX Spark Systems Can Work Together
NVIDIA is also introducing a way to connect two 64GB DGX Spark systems into a larger local AI environment.
Using NVIDIA Sync Cluster Assistant, two systems can be connected through their networking hardware to create a combined memory pool of up to 128GB.
This can allow larger AI models and more demanding workloads to be distributed across two machines.
NVIDIA says two connected 64GB systems can support models of up to approximately 200 billion parameters.
NVIDIA Reports Up to 1.7x Performance in a Test
The company says its testing with the Qwen 3.8 27B model showed that two clustered 64GB DGX Spark systems delivered up to 1.7 times the performance of a single system.
Actual performance can vary depending on the model, workload, software configuration and other factors.
The clustering capability is designed to allow developers to start with one machine and expand their local AI environment as their requirements increase.
AI Software Comes Built In
The DGX Spark platform is designed to work with NVIDIA's broader AI software ecosystem.
The system supports tools and frameworks including PyTorch, vLLM, Ollama and other AI development software.
NVIDIA also provides its own AI libraries, models and development tools, allowing developers to begin experimenting with AI workloads without building the entire software environment from scratch.
Major PC Manufacturers Will Offer the New System
The 64GB DGX Spark configuration will be available through several major computer manufacturers.
NVIDIA has announced availability through Acer, ASUS, Dell, Gigabyte, HP and MSI.
This expands the number of hardware manufacturers involved in NVIDIA's local AI computing ecosystem and gives developers more options when purchasing the system.
Why Local AI Is Becoming Important
Cloud computing has become the primary infrastructure for many advanced AI services, but local AI is gaining attention for several reasons.
Running models locally can reduce dependence on internet connectivity and can allow sensitive data to remain on a company's own hardware.
For developers, local AI can also provide a convenient environment for testing models, applications and agents without repeatedly paying for remote computing resources.
The growth of smaller and more efficient AI models is making local AI increasingly practical on powerful desktop systems.
AI Hardware Costs Are Also Changing
The launch comes during a period of strong demand for memory and computing hardware used in AI systems.
Independent technology reporting has noted significant price changes across the DGX Spark product range, with the existing 128GB configuration becoming considerably more expensive than its original launch price.
The new 64GB configuration gives NVIDIA another price point for developers who do not require the memory capacity of the larger system.
However, at $4,999, the DGX Spark remains a specialized computing product aimed primarily at AI developers, researchers and professional users rather than ordinary desktop consumers.
From Cloud AI to Personal AI Supercomputers
The development of compact AI computers represents a broader shift in the artificial intelligence industry.
Large AI models traditionally require massive data-center infrastructure containing thousands of GPUs or other specialized processors.
At the same time, advances in model efficiency and AI hardware are allowing increasingly capable models to operate on smaller systems.
This creates a new category between ordinary personal computers and massive cloud data centers: powerful local machines specifically designed for AI workloads.
What Comes Next for DGX Spark?
NVIDIA plans to make the new 64GB DGX Spark available from October 23 through its hardware partners, starting at $4,999.
The company is also developing additional software tools that make it easier to launch AI models and connect multiple systems.
For developers, the ability to run sophisticated AI models locally could make experimentation faster and give organizations more control over their data and computing infrastructure.
As AI agents become more capable and AI models continue to become smaller and more efficient, local AI computers such as DGX Spark could become an increasingly important part of the AI development ecosystem.
The Local AI Race Is Getting Bigger
NVIDIA's new DGX Spark configuration shows how the AI industry is expanding beyond massive cloud data centers.
The combination of the Grace Blackwell platform, 64GB unified memory, local AI software and multi-system clustering gives developers a dedicated environment for experimenting with advanced AI workloads.
With availability beginning later this month, the new system will give developers another way to build and operate AI applications without relying exclusively on cloud infrastructure.
The next phase of AI computing may therefore involve both enormous data centers and increasingly powerful machines sitting directly on developers' desks.
Journalist: Vijay Singh