Abstract: With the fast development of numerical weather prediction (NWP) technology, the quantitative precipitation forecast of NWP models has become a major basis of everyday operation weather forecast in most countries all over the world at the moment. The precipitation forecasts of five NWP models, including T213L31, HLAFS 0.25, Huabei-meso-MM5, a Germen global model and Japanese global model which are widely applied in National Meteorological Center and other weather prediction offices in China, are verified and assessed synthetically by the spatial distribution of seasonal mean daily rain rate, the time series of regional mean rainfall rate and statistic verification method. The result of the verification shows that there is certain spatial forecast capability for the quantitative precipitation forecasting of short term (for forecast durations of 24/36~48/60 h at 24-hour intervals) by NWP models at present, but there are some biases or errors for position of heavy rainfall center or rain belt. With the forecasting valid time increasing, the errors of rain belt increase. Most models often overestimate the rainfall in the western part of China or underestimate the precipitation in the eastern part of China. Through the verification of the time series of regional mean daily precipitation, it is found that there is more obvious forecasting capability of NWP model for the developing trends of heavy rainfall processes in some regions in the eastern part of China than ordinary processes, and the forecast of the middle and lower reaches of the Yangtze is the best. However, there are some errors for the quantity of forecast precipitation against the observation rainfall. For example, MM5 model often overestimates the rainfall in South China while HLAFS model usually overestimates the precipitation in Southwest China. According to the results of accumulative verification, there are not much differences for Ts of light rain and moderate rain among five models. The global models' Ts are higher than those of regional models, and the forecast by Japanese model is the best among the five models forecast for the two classes. But for heavy rain or more severe rain, the performance of domestic models is super to foreign models. There are unique advantages of MM5 and T213L31 model for different valid time or rainfall classification, but for quantity and area of forecast rainfall are greater than those of observation.