Objective: The widespread adoption of building information modeling (BIM) in construction projects has notably advanced design, coordination, and project management. However, a persistent challenge in BIM implementation lies in the efficient management of geometric information, particularly when identical or similar components are repeatedly present in a model. This leads to substantial data redundancy, increasing both storage and network transmission costs, thereby impeding the scalability and efficiency of BIM models in large-scale projects. Methods: This study proposes an approach for geometric component comparison and reuse based on geometric tensor analysis and graph matching algorithms. The method focuses on BIM models in the industry foundation class (IFC) format, leveraging the boundary representation (B-Rep) of geometric components to extract key features such as metric tensors and inertia tensors. These tensors serve as shape and spatial property descriptors, enabling accurate assessments of geometric similarity. Graphs are constructed in this process, where nodes represent the surfaces of components and edges denote the topological relationships between them. By applying graph matching techniques, this method identifies geometric similarity even when components undergo transformations such as rotation, translation, or scaling. The proposed method was validated through several experiments focused on reducing geometric redundancy and optimizing model storage. Results: In the first experiment, a complex geometric component, a window that consisted of 392 surfaces and 758 edges, was analyzed for reuse. This process reduced the IFC file size from 188 kB to 66 kB, representing a 64.9% decrease and demonstrating the effectiveness of identifying and reusing repeated geometric components in minimizing storage requirements. The second experiment applied the method to a 22-story residential building, focusing on the standard floors comprising 22 718 components. The method decreased the overall IFC model size by 90.0%, illustrating its scalability and efficiency in handling large-scale BIM models. The third experiment evaluated the time required to retrieve models of different scales, showing short retrieval times. However, these findings indicated the need for further optimization when handling large-scale models. This approach offers several advantages over traditional methods for managing geometric information in BIM. First, geometric tensor analysis ensures robust component comparison that is invariant to transformations, enabling accurate identification and reuse of components regardless of their spatial orientation. Second, the integration of graph matching algorithms provides a flexible and scalable framework for handling complex topologies and large datasets, making this method particularly suitable for high-volume construction projects. In addition to reducing redundant geometric data in BIM models, the proposed method enhanced storage efficiency and data exchange, minimized BIM model sizes, and facilitated faster data transfer while reducing the strain on network resources. These features are critical for large collaborative projects involving multiple stakeholders across different platforms. Furthermore, the method improved the maintainability of BIM models by facilitating version control and consistency checks, ensuring efficient management of model updates without duplicating existing geometric data. Conclusions: In conclusion, combining geometric tensor analysis and graph matching offers a robust solution for optimizing BIM models through geometric component reuse. This method addresses the challenge of data redundancy, delivering significant improvements in storage efficiency and network transmission. Subsequent research may focus on refining the computational aspects of the method, particularly for processing highly complex models, and explore its integration into cloud-based BIM systems to further enhance real-time collaboration and model management across multidisciplinary teams.