This study focuses on investigating the deviation of 2006-2013 mean precipitation projected by the multi-model ensemble over China using the outputs of 10 CMIP5 climate models under various RCP scenarios, in comparison with the latest grid precipitation from the CRU, University of East Anglia. The results show that there are obvious differences in projected precipitation among the 10 models for Northwest China and the eastern coastal zone. The model-ensemble precipitation is overestimated in northern and western parts of China and underestimated in the coastal zone. In the winter-half year, the precipitation is obviously overestimated in most parts of China, even double or more than observed, but underestimated in the coastal zone; while in the summer-half year, it is underestimated in the eastern monsoon zone and overestimated in western China. The deviation changes with time, and its trend is apparent in northern and eastern parts in the summer-half year, and in Northeast, eastern and southern China in the winter-half year. In addition, the overestimation of precipitation is intensified in the northwest of China if using RCP8.5 scenario, but relatively weaker for the scenarios in East or South China. Besides, although the patterns of deviation are similar in El Niño and La Niña years, the interannual difference of the deviation is still obvious in some parts of North China and coastal areas. The different deviation seemingly results from defects in the models themselves, such as their physical processes, cumulus convection parameterizations, representation of solid precipitation, topography treatment, and spatial resolution. The deviation reveals that there would be severe uncertainty if estimating China precipitation directly using CMIP5 model outputs. Hence, it is necessary to estimate the uncertainty before using the output data, so as to reduce the potential risk for users or decision-makers responsible for future development plans.