Climatology of the observed daily precipitation extreme indices (SDII, simple daily intensity index; CDD, the maximum number of consecutive dry days; R10, number of days with precipitation greater than 10 mm; R5d, maximum 5-day precipitation total; R95t, fraction of total precipitation due to events exceeding the 95th percentile of the climatological distribution for wet day amounts) at 550 stations in China during 1961－2000 is used to evaluate the simulation ability of 7 IPCC AR4 Coupled Climate Models, the projected change of the precipitation extreme indices over China under IPCC SRES A2、A1B and B1 is also studied. The results show that the state-of-the-art IPCC AR4 models can simulate the spatial distributions and the linear trends of precipitation extremes well. The multi-model ensemble (MME) shows the best skill, but both the MME and single model fail to simulate the interannual variability and have large biases, such as there are excessive extreme precipitation over the eastern side of the Tibetan Plateau while in the monsoon regions the modeled intensity of precipitation extremes is lower than the observation. In the 21st century the precipitation will become more “extreme”， there would be an overall increasing trend in extreme precipitation events over most of China under a warming environment, and the change scope is scaled to the emissions scenarios.