The ability of climate models in reproducing climate over the Qinghai-Tibetan Plateau, where the natural environment is sensitive and vulnerable to climate change, remains unclear. Here, we examine the performance of 44 models participating in the Fifth Phase of the Coupled Model Intercomparison Project (CMIP5) over the Qinghai-Tibetan Plateau by comparing their outputs with ground observations of surface air temperature and precipitation for the period 1986-2005. The results show that CMIP5 models underestimate annual and seasonal temperatures, with an average of -2.3℃ for the annual mean and larger cold biases in autumn and winter. CMIP5 models can reasonably capture the climatological spatial patterns of annual and seasonal temperatures, but they overestimate spatial variability. The ability of CMIP5 models in reproducing annual and seasonal temperatures can be improved through topographic correction. Comparatively, CMIP5 models perform poorly in reproducing annual and seasonal precipitation. They can capture the climatological spatial pattern of annual precipitation that mainly features a northwest-to-southeast increase, but they overestimate annual and seasonal precipitation amounts, with an average of 1.3 mm d-1 for the annual mean mainly derived from spring and summer. Moreover, the simulated spatial variability of annual and seasonal precipitation is greater than that in the observation, particularly in spring and winter. In general, the ensemble mean of 44 models shows a better skill than most of individual models; the 44-model ensemble mean with the same weights performs better than the median of 44 models; and the ensemble mean of the chosen models with a demonstrable ability can further improve the skills of climate models, particularly for annual and seasonal precipitation.