Description
The NVIDIA DGX Spark is a next-generation AI workstation engineered to accelerate the development, testing and deployment of advanced artificial intelligence models. Built around the NVIDIA GB10 Grace Blackwell Superchip, this compact powerhouse delivers up to one petaFLOP of FP4 AI performance, providing developers with the computational power required to build and refine the latest generation of AI models.
With 128 GB of coherent unified system memory, DGX Spark allows developers to work locally with AI models containing up to 200 billion parameters. This capability enables rapid prototyping, model fine-tuning and high-performance inference without relying on external data centres or cloud infrastructure.
The system comes preconfigured with the NVIDIA AI software stack, giving users access to a comprehensive ecosystem of frameworks, tools and libraries optimised for generative AI workloads. Developers can immediately begin building solutions using models from leading AI providers including DeepSeek, Meta, NVIDIA, Google, Qwen and more.
High-performance NVIDIA ConnectX networking also allows multiple DGX Spark systems to be linked together, enabling teams to scale their development environment and work with AI models of up to 405 billion parameters.
Despite its immense computing capability, the DGX Spark maintains a compact, power-efficient desktop form factor, making it ideal for AI developers, researchers and data scientists who require supercomputer-level performance in a local workstation environment.
From prototyping and experimentation through to deployment and edge application development, the NVIDIA DGX Spark delivers the tools and performance needed to accelerate modern AI innovation.
Key Features:
- NVIDIA GB10 Grace Blackwell Superchip delivering up to 1 petaFLOP of FP4 AI performance
- 128 GB coherent unified system memory for running large-scale AI models locally
- Supports AI models with up to 200 billion parameters
- NVIDIA ConnectX networking enables two systems to work with models up to 405 billion parameters
- Preinstalled NVIDIA AI software stack with tools, frameworks, libraries and NVIDIA NIM
- Ideal for AI development, machine learning and generative AI workloads
- Optimised for prototyping, fine-tuning and inference
- Supports edge AI development using frameworks including Isaac™ and Metropolis
- Compact, power-efficient desktop design delivering AI supercomputer performance
