NVIDIA DGX Station: A Leap Forward for Local AI Model Development

Back in October, we highlighted how Docker Model Runner on the NVIDIA DGX Spark simplified local AI model development, offering a familiar Docker experience to run large models. This solution proved popular among developers who could now replace complex GPU setups and cloud API calls with ease.

The recent unveiling of NVIDIA DGX Station at NVIDIA GTC 2026 has further advanced this trend. This new workstation, built around the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, promises to bring even more power and flexibility to local AI development. To support this, Docker has announced that Model Runner will now be compatible with the DGX Station.

From NVIDIA DGX Spark to DGX Station: What Has Changed?

The NVIDIA DGX Spark, introduced earlier, offered a compact and powerful solution with 128GB of unified memory and petaflop-class AI performance. It was an excellent entry point for developers looking to run complex models on their local machines.

However, the NVIDIA DGX Station represents a significant leap forward in terms of both hardware capabilities and user experience. This workstation is built around the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, which integrates 72-core NVIDIA Grace CPUs with high-performance NVIDIA Blackwell Ultra GPUs through NVLink-C2C technology. The result is an incredibly powerful system that can handle even the most demanding AI tasks.

With Docker Model Runner support for DGX Station, developers now have a seamless and intuitive way to run large models locally. This compatibility ensures that users familiar with Docker’s ecosystem will find it easy to transition from their current workflows to this new hardware setup without any significant learning curve.