DGX Spark provides a complete local platform to develop, test, and deploy autonomous AI agents. With the NVIDIA AI software stack preinstalled—including tools for agent orchestration, evaluation, and secure local deployment—developers can build agents that run persistently and privately, without cloud dependency. From single-agent prototypes to multi-agent workflows, DGX Spark supports the full agentic development cycle.
NVIDIA DGX Spark
Designed to build and run autonomous agents.
Where to Buy
Find your best purchase option on the NVIDIA Marketplace.
A New Beginning for Windows PCs
Powered by NVIDIA RTX Spark.
Overview
Your Desktop Agent Computer
Powered by the NVIDIA GB10 Grace Blackwell Superchip, NVIDIA DGX Spark™ is a complete platform for local autonomous agents. Large local memory, powerful AI compute and the NVIDIA AI software stack enable agents and large models to run locally, reducing the need for cloud-based token generation resources. With a compact, power-efficient design, DGX Spark is built to run always-on agent workloads—right from your desktop.
NVIDIA GB10 Grace Blackwell Superchip
Up to
20-core
Grace CPU
Up to
1 petaFLOP
FP4 AI Performance
Up to
128 GB
Unified Memory
NVIDIA GB10 Superchip
Experience up to 1 petaFLOP of AI performance at FP4 precision with the NVIDIA Grace Blackwell architecture.
Coherent Unified System Memory
Run AI development and testing workloads with state-of-the-art AI models at your desktop with a large, unified system memory. Available in 64 GB* or 128 GB memory configuration options.
NVIDIA ConnectX Networking
High-performance NVIDIA ConnectX™ networking enables up to four DGX Spark systems to work together on larger AI models, deliver faster inference, and run multi-agent workloads.
NVIDIA AI Software Stack
Utilize a full-stack solution for generative AI workloads, encompassing NVIDIA tools, frameworks, libraries, and pre-trained models, including NVIDIA NIM™.
NVIDIA Sync
Connect to your DGX Spark from your laptop or desktop and monitor system status, launch Cursor, VS Code, or AI Workbench remotely, and configure up to four devices into a high-speed multi-node cluster. Install the latest software updates to get the most out of your system.
*64 GB memory configuration is available exclusively through participating OEM partners.
Workloads
Accelerate All Your AI Workloads
Delivering the power of Grace Blackwell in a desktop-friendly size, DGX Spark is ideal for AI developer, researcher, and data scientist workloads.
Agentic AI
Build, run, and manage always-on AI agents from your desk.
Prototyping
Develop, test, and validate AI models and applications.
Fine-Tuning
Fine-tune AI models up to 70 billion parameters.
Inference
Test, validate, and inference with AI models up to 200 billion parameters.
Data Science
High-performance data science at your desk.
Edge Applications
Develop edge applications with NVIDIA AI frameworks, including NVIDIA Isaac™, Metropolis, and many others.
Scale to Larger Models, Faster Inference, and Agentic Workloads
Run frontier-class models on desktop clusters.
*64 GB memory configuration is available exclusively through participating OEM partners.
DGX Spark for Developers
Everything You Need to Build, Run, and Deploy AI Locally
Playbooks for Everyone
From novice to expert, leverage our curated playbooks to get off to a running start on various projects. Playbooks contain step-by-step recipes to demonstrate the art of the possible.
Resources
Explore NVIDIA DGX Spark
-
Blogs
-
Videos
-
Livestreams
Partners
NVIDIA DGX Spark
Get DGX Spark through our authorized channel and retail partners.
NVIDIA DGX Spark Powered Systems
Discover NVIDIA GB10 Grace Blackwell Superchip–powered systems from our preferred OEM partners.
FAQ
NVIDIA DGX Spark belongs to a new class of computers designed from the ground up to build and run AI. Powered by the NVIDIA GB10 Grace Blackwell Superchip, it's a complete local AI platform for agents, LLM, image and video, and multimodal inference, fine-tuning, data science, and edge development, bringing the power of Grace Blackwell to every desk. Connect up to four systems via NVIDIA ConnectX-7 to scale to larger models and more complex workflows
NVIDIA DGX Spark is designed for AI developers, researchers, data scientists, and students who need powerful, always-on AI compute on the desktop—without relying on cloud or data center access for every workload. Whether working independently or as part of an enterprise fleet, DGX Spark users share a common need: large local memory, the full NVIDIA AI software stack, and a development environment consistent with production infrastructure.
DGX Spark ships with DGX OS and additional NVIDIA tools. DGX OS is a Linux distribution based on Ubuntu and matches the Linux distribution running on other DGX platforms. DGX Spark comes pre-installed and/or supports a large catalog of tools, frameworks, models, and software for development use from NVIDIA NGC. For example:
- NVIDIA NeMo framework for model fine-tuning
- NVIDIA RAPIDS libraries for accelerated data science
- NVIDIA Blueprints for building agentic AI applications
- NVIDIA NIM microservices for testing optimized inference results
- NVIDIA NemoClaw, an open-source reference stack that adds privacy and security to OpenClaw, with a single-command install of the OpenShell runtime, NVIDIA Agent Toolkit, and Nemotron models
- Common frameworks such as PyTorch, Python, and Jupyter notebooks
DGX Spark users can register free of charge for the NVIDIA Developer Program to get access to NGC including additional SDKs, toolkits, trained models, and workflows.
Hardware support will be provided by the vendor that sold the system.
Up to 4 DGX Spark systems can be connected using NVIDIA ConnectX-7 networking. You can cluster two or three systems with a single QSFP cable; clustering 4 systems requires an external network switch. For optimal load balance and parallelism, connect DGX Spark in a power-of-2 number
| Configuration | DGX Spark 64 GB* | DGX Spark 128 GB | Connection |
|---|---|---|---|
| 1 × DGX Spark | 64 GB | 128 GB | – |
| 2 × DGX Spark | 128 GB | 256 GB | Direct QSFP cable, no switch required |
| 4 × DGX Spark | 256 GB | 512 GB | 200 GbE switch (QSFP56-DD, RoCE v2) |
Details on QSFP Cables and Switches
Each DGX Spark has two QSFP ports on the back of the device. Each port provides up to 200 Gigabits per second (Gb/s), but the incoming speed also depends on the cable you use. Approved cables are:
Amphenol: NJAAKK-N911 (QSFP to QSFP112, 32AWG, 400mm, LSZH), NJAAKK0006 is the 0.5m version of this cable
Luxshare: LMTQF022-SD-R (QSFP112 400G DAC Cable, 400mm, 30AWG)
Minimum requirements for switch:
- Minimum 4 ports QSFP56-DD
- Breakout to 25/50/100/200/400 G
- Recommended max port speed of 200G–400G per port
- One 1/10GbE management Ethernet port
- Supports RoCE v2
- Switching capacity/throughput: minimum 0.8–1.6 Tbps
*64 GB memory configuration is available exclusively through participating OEM partners.
The enterprise manageability features developed for DGX Spark 128 GB can also be applied to DGX Spark 64 GB. Full details are located at https://docs.nvidia.com/dgx/dgx-spark/enterprise-manageability.html