| Based on integrated simulations of 26 global climate models provided by the Coupled Model Intercomparison Project (CMIP), this study predicts changes in temperature and precipitation across China in the 21st century under different representative concentration pathways (RCPs), and analyzes uncertainties of the predictions using Taylor diagrams. Results show that increases of average annual temperature in China using three RCPs (RCP2.6, RCP4.5, RCP8.5) are 1.87 °C, 2.88 °C and 5.51 °C, respectively. Increases in average annual precipitation are 0.124, 0.214, and 0.323 mm/day, respectively. The increased temperature and precipitation in the 21st century are mainly contributed by the Tibetan Plateau and Northeast China. Uncertainty analysis shows that most CMIP5 models could predict temperature well, but had a relatively large deviation in predicting precipitation in China in the 21st century. Deviation analysis shows that more than 80% of the area of China had stronger signals than noise for temperature prediction; however, the area proportion that had meaningful signals for precipitation prediction was less than 20%. Thus, the multi-model ensemble was more reliable in predicting temperature than precipitation because of large uncertainties of precipitation.