A compact, affordable development system with a high-memory GPU that allows you to work with large AI models - all with the full NVIDIA AI software and subscription package.

The power of artificial intelligence in your hands

As generative AI models become larger and more complex, it becomes more and more difficult to develop them locally on your own machine. Prototyping, fine-tuning and running requires a lot of memory and high computing power. As companies, software developers, government agencies, startups and researchers work more and more intensively on AI developments, the demand for suitable hardware is constantly increasing. However, traditional workstations are often no longer able to handle these resource-intensive tasks, so new solutions are needed.

NVIDIA DGX™ Spark offers a solution for this, which works with the new NVIDIA GB10 Grace Blackwell superchip. This device delivers an incredible 1 petaFLOP of AI performance in a compact, energy-efficient form factor. With the pre-installed NVIDIA AI software suite and 128 GB of memory, developers can work with the latest models with up to 200 billion parameters locally - such as models from DeepSeek, Meta or Google - and then seamlessly transfer them to a data center or cloud.

NVIDIA DGX Spark - highlights

NVIDIA GB10 superchip

Based on the Grace Blackwell architecture, the chip delivers up to 1 petaFLOP of AI performance with FP4 accuracy – a huge computing power for artificial intelligence tasks.

128 GB unified system memory

Thanks to its high-capacity, uniform memory, you can work with AI models with up to 200 billion parameters directly on your desktop - ideal for development and testing.

NVIDIA ConnectX network

The high-performance ConnectX network technology enables you to manage even larger models – up to 405 billion parameters – by connecting two DGX Spark systems.

Complete NVIDIA AI software environment

You get a complete solution for generative AI: it includes all the necessary tools, libraries, frameworks and pre-trained models that you might need for your work.

Get involved in our development!

The DGX Spark reference model is already available in NVIDIA's range, and soon you will find it in the product range of manufacturers such as ASUS or Dell Technologies.

Do you want even more performance? Request an offer for two NVIDIA DGX Spark units, which can be directly connected using the built-in ConnectX-7 SmartNIC, so models with up to 405 billion parameters can be run.

- Architecture NVIDIA Grace Blackwell
- GPU Blackwell architecture
- CPU 20 core Arm, 10 Cortex-X925 + 10 Cortex-A725 Arm
- CUDA cores Blackwell generation
- Tensor cores 5th generation
- RT cores 4th generation
- Tensor performance1 1 PFLOP
- System memory 128 GB LPDDR5x, coherent unified system memory
- Memory interface 256 bits
- Memory bandwidth 273 GB/s
- Storage 4 TB NVME.M2 with self-encryption
- USB 4x USB Type-C
- Ethernet 1x RJ-45 connector
- 10 GbE
- NIC ConnectX-7 NIC @ 200 Gbps
- Wi-Fi Wi-Fi 7
- Bluetooth BT 5.4
- Audio output HDMI multi-channel audio output
- Power supply 240 W
- GB10 TDP2 140W
- Display connectors 1x HDMI 2.1a
- NVENC | NVDEC 1x | 1x
- Operating system: NVIDIA DGX™ operating system
- System dimensions: 150 mm L x 150 mm W x 50.5 mm H
- System weight: 1.2 kg