Abstract: Spatial variability is a unique property of geo-materials, posing significant challenges to geotechnical reliability analysis and risk assessment. Within the framework of random finite element method, this study proposes a collaborative risk assessment approach that combines the advantages of multiple models in terms of efficiency and accuracy. By establishing an inherent connection between simple and complex models through the sample space of stochastic simulation, the proposed approach not only flexibly handles the complex spatial variability of geo-materials but also achieves both efficient and accurate risk assessment. The effectiveness of the proposed approach is demonstrated through three typical geotechnical examples of excavation, shallow foundation, and slope. Its computational efficiency is improved by several orders of magnitude compared to conventional random simulation methods, providing a practical tool for quantitative risk assessment in complex engineering projects. The three-dimensional spatial variability significantly affects the failure modes, reliability, and risk of geotechnical structures. The collaborative risk assessment approach can appropriately consider these effects and greatly facilitate the application of random finite element method in practice. The multi-model collaboration strategy can be extended to various stochastic problems with well-defined objectives, such as Bayesian updating and optimization.