Heterogeneous lattice metamaterials exhibit superior mechanical properties and great design flexibility compared to single-type lattice metamaterials. However, it is still challenging to simultaneously optimize the distribution of topology configurations and volume fractions for heterogeneous lattice metamaterials. In this study, the hybrid metamaterials are generated by combing the cross-shaped and X-shaped lattice metamaterials, and the continuous transitions between topology configurations are realized. Mechanical properties and deformation behaviors of these lattice metamaterials are systematically investigated, and the results indicate that the deformation modes of lattice metamaterials depend on both topology configurations and loading conditions. Moreover, a surrogate model is developed to predict the mechanical properties of lattice metamaterials, which eliminates need for time-consuming homogenization calculations during multiscale optimization process. A multiscale optimization framework for heterogeneous lattice metamaterials is then established and validated through a cantilever beam case study. The proposed method achieves simultaneous optimization of volume fraction and topology configuration, reducing the compliance of cantilever beam by approximately 50.3%. Compared to traditional multiscale optimization methods, the optimized heterogeneous lattice metamaterials exhibit an additional compliance reduction of 0.8–45.7%. The proposed method provides a promising strategy for designing high-performance lightweight components for aerospace and automobile applications.