Abstract: To address the problems that railway engineering BIM (Building Information Modeling) models have large data volume and high geometric complexity, which cause slow cloud loading and operation lag, this paper investigated key technologies for rendering enhancement and lightweighting of railway engineering BIM models. This study adopted Three.js to construct 3D visualization scenes and data adapted to Railworks component documents and assembly documents. It studied the lightweight mechanism based on the quadratic error metric algorithm and the data-model separation strategy to reduce the massive number of meshes in railway engineering BIM models. Meanwhile, it applied a dynamic loading strategy to resolve operation lag for users. Demonstration applications in the Jintang Subsea Tunnel and the Xihoumen Rail-cum-Road Bridge of the Ningbo-Zhoushan Railway showed that the study improves the BIM model loading speed by more than 60%, realizes efficient web-based visualization, and promotes the large-scale application of domestic BIM technologies in cloud collaborative environments.