In order to improve the ability of T639 model, the synoptic verification on its mediumrange forecasting in autumn is made in comparison with ECMWF and Japan models. The results show that the three models have good performances in the aspect of predicting the largescale circulation evolution and adjustment in the middle and high latitudes of Asian areas， but the T639 model shows a relative weak performance advanced 144 hours. The T639 model failed in forecasting the track and the JP model underestimated the intensity of FANAPI (No.1011). Compared with the T639 model and Japan model, the ECMWF model is better in forecasting most weather systems and gets the most similar to the observation.