In order to understand the forecast skills of subseasonal to seasonal (S2S) models, the summer daily precipitation and extreme precipitation forecasts from China Meteorological Administration (CMA), European Centre for Medium-Range Weather Forecasts (ECMWF), and National Centers for Environmental Prediction (NCEP) models during their common hindcast period from 1999 to 2010 are evaluated by using the precipitation observation data from stations in the eastern part of Chinese mainland east of 108°E. The results show that ECMWF has the best overall forecast performance, followed by NCEP, and CMA is relatively worse. With the increase of forecast time, each model shows a characteristic that the forecast tends to be dry (wet) when the observation is wet (dry). The forecast skills of the models are almost lost on the S2S time scale, so there is a lot of room for improvement. The evaluation results of a single station are directly influenced by the definition method of the critical threshold of extreme precipitation, which has little effect on the overall forecast skills of the region. With the increase of observed precipitation, the root mean square error of S2S model forecast tends to increase. The correlation coefficient between forecast and observation shows a continuous (oscillatory) decrease or even negative correlation in all (extreme) precipitation events. The mean square skill score is higher in the case of more precipitation events. In all precipitation events, the false alarm rate of each model is much higher than the missing alarm rate, but it is opposite in extreme precipitation events. The performance of precipitation forecast verification indices in absolute extreme precipitation classification verification becomes worse with the increase of precipitation grade. There is almost no severe rainstorm forecast in all models, and the forecast accuracy of CMA model on the occurrence of extreme precipitation events is poor.