The 3D Gaussian Splatting method for 3D environment reconstruction from images brought significant advancements to photorealistic novel-view synthesis. It combines the advantages of primitive-based rendering with a differentiable renderer, thus obtaining state-of-the-art image quality and surpassing neural methods for scene representation in optimization and rendering speed. This is a significant step towards bringing these methods to real-time consumer applications, however, 3DGS still requires significant computing power which is not available in consumer devices. In this project, I will present a method for accelerating 3DGS rendering through a hierarchical Level of Detail structure that combines a regular octree subdivision with feature-based primitive clustering to obtain lower-detail representations. Also, I will present a level selection solution to maintain the desired detail granularity across the scene by computing a dynamic cut through the scene tree representation. This method achieves a reduction in the frame time between 14% and 33% by reducing the number of primitives in the scene to around 50%, a reduction which maintains the image quality above 31 dB PSNR compared to the original reconstruction.