CityGaussian (CityGS) is a real-time large-scale scene rendering method that employs a divide-and-conquer training strategy and Level-of-Detail (LoD) strategy to efficiently train and render 3D Gaussian Splatting (3DGS). The method addresses the challenges of training and rendering large-scale scenes by partitioning the scene into blocks, using a global Gaussian prior for efficient training, and dynamically selecting detail levels based on distance from the camera. CityGS achieves state-of-the-art rendering quality and enables real-time rendering across vastly different scales. The method is evaluated on large-scale scenes, demonstrating superior performance in terms of rendering fidelity and speed. Key contributions include an effective divide-and-conquer strategy for parallel training, a LoD strategy for real-time rendering under varying scales, and a method that performs favorably against current state-of-the-art approaches. The work is supported by multiple research institutions and is available on a project page for further exploration.CityGaussian (CityGS) is a real-time large-scale scene rendering method that employs a divide-and-conquer training strategy and Level-of-Detail (LoD) strategy to efficiently train and render 3D Gaussian Splatting (3DGS). The method addresses the challenges of training and rendering large-scale scenes by partitioning the scene into blocks, using a global Gaussian prior for efficient training, and dynamically selecting detail levels based on distance from the camera. CityGS achieves state-of-the-art rendering quality and enables real-time rendering across vastly different scales. The method is evaluated on large-scale scenes, demonstrating superior performance in terms of rendering fidelity and speed. Key contributions include an effective divide-and-conquer strategy for parallel training, a LoD strategy for real-time rendering under varying scales, and a method that performs favorably against current state-of-the-art approaches. The work is supported by multiple research institutions and is available on a project page for further exploration.