A realistic simulation of rainfall over the Tibetan Plateau (TP) is a big challenge for both regional and global climate models. In this study, we investigate the simulations of summer rainfall over the TP from 1979 to 2010 by the Weather Research and Forecasting (WRF) model with various horizontal resolutions and cumulus schemes, with a focus on the difference in the model's skill in simulating interannual variability of rainfall between early and high summer. The WRF captures spatial pattern of climatological summer mean rainfall over the TP. However, it produces apparent wet bias, especially in southern and eastern edges of the TP. Despite this climatological bias, WRF skilfully reproduces the interannual variability of summer rainfall over the southeastern TP, where maximum rainfall is located. An increase in horizontal resolution or an appropriate cumulus scheme mostly improves the simulation of climatological mean rainfall. However, the phase of interannual variability of simulated rainfall is not sensitive to horizontal resolution and cumulus scheme. Instead, it is sensitive to mechanisms responsible for interannual variability of rainfall. WRF has a high skill in simulating the interannual variability of rainfall over the southeastern TP in July and August, but a low skill in June. The WRF's high skill is attributed to that the interannual variability of July–August rainfall is largely driven by large‐scale circulation, while its low skill for June rainfall may be ascribed to the overestimation of snow and its relationship with rainfall.