NVIDIA introduces DGX Spark and DGX Station to accelerate innovation with open-source AI models, offering high performance for developers. The systems support running advanced AI models from desktops, providing new capabilities to easily scale to the cloud. The NVIDIA Blackwell architecture powers DGX Spark, enabling compression of AI models by up to 70%.

DGX Spark and DGX Station offer powerful AI capabilities for developers, researchers, and data scientists, allowing them to build, fine-tune, and run AI models locally. The systems come preconfigured with NVIDIA AI software and CUDA-X libraries, providing plug-and-play optimization. NVIDIA’s collaborations with open-source software ecosystem enhance performance by 35% on average.

DGX Station, with the GB300 Grace Blackwell Ultra superchip and 775GB of coherent memory, can run models up to 1 trillion parameters, including advanced AI models like Kimi-K2 Thinking and OpenAI gpt-oss-120b. The compact form factor of DGX Station enables projects to test and develop directly on powerful GB300 superchip, accelerating development cycles.

Industry leaders, software innovators, and global partners are adopting DGX Spark for local inference and AI workflows. DGX Spark enables faster iteration, greater data control, and interactive AI experiences on the desktop. At CES, NVIDIA demonstrates how DGX Spark powers AI agents like the Hugging Face Reachy Mini robot.

DGX Spark playbooks help developers kickstart real-world AI projects, with new additions covering topics like robotics training, vision language models, and financial analysis. NVIDIA AI Enterprise software support is now available for DGX Spark and GB10 systems, enabling fast and reliable AI engineering and deployment. Licenses are expected to be available at the end of January.

DGX Spark and DGX Station are available from various manufacturer partners, including Acer, ASUS, Dell Technologies, HP Inc., Lenovo, MSI, and PNY. DGX Station will be available starting in spring 2026 from ASUS, Boxx, Dell Technologies, MSI, and Supermicro. Explore DGX Spark further in a technical blog provided by NVIDIA.

Read more at NVIDIA: NVIDIA DGX Spark and DGX Station Power the Latest Open-Source and Frontier Models From the Desktop