The raw-material mix ratio and preparation of similar materials are crucial for the success of physical model tests and for accurately reflecting prototype properties. In this study, an optimum similar material for plateau alluvial and lacustrine (PAL) round gravel was developed based on similarity theory. The similar materials were subjected to sensitivity factor analysis and microscopic analysis. Subsequently, the optimum similar material was applied to a three-dimensional (3D) physical model test of an ultradeep foundation pit (FP). The findings show that the similar material prepared with gypsum, LD, bentonite, water, barite powder, and DS at a ratio of 1:1:1.4:3.5:8.8:13.2 was the best for a 3D physical model test of the ultradeep FP in PAL round gravel strata. The sensitivity-factor analysis revealed that barite powder had the greatest impact on gamma, that c and phi were primarily affected by bentonite, and that the LD-gypsum ratio controlled E. A nonuniform particle-size distribution as well as the presence of edge-to-face contacts and small pores between particles constituted the microphysical factors affecting the mechanical properties of the optimum similar material. Using dolomite with a Mohs hardness of 3.5-4 instead of traditional quartz sand with a Mohs hardness of 7 as the raw material can produce a similar material for the target soil with mechanical parameters closer to those of the ideal similar material. The application of the optimum similar material in physical model tests has revealed the stress field response law of ultra deep foundation pit excavation. This study could provide reference and inspiration for the development of similar materials in gravel formations with weaker mechanical properties.