With the widespread adoption of remote work during the COVID-19 pandemic, metaverse technologies have expanded beyond gaming into business environments. The spread of smartphones with depth-sensing capabilities has made 3D scanning and modeling of physical objects more accessible, leading to increased transmission of 3D models over networks. While improved network infrastructure can handle large-volume 3D models, using high-fidelity scanned models directly in metaverse environments can decrease system responsiveness. One solution is to simplify and optimize 3D models by approximating the original shape using basic geometric bricks, similar to building bricks. In this paper, we propose a method for approximating scanned 3D models using a small number of basic bricks. While our previous research using LEGO bricks in a layered structure showed promise, approximation accuracy was limited by brick constraints based on the number of layers. We address this issue by providing bricks of different thicknesses. Our approach achieves high approximation accuracy while reducing the number of bricks by applying a packing method that uses bricks of different heights for internal and external model structures. This reduces data transfer volumes and computational load in metaverse environments. We evaluate our method’s effectiveness using various 3D model datasets.