The performances of mediumrange forecasts are verified and compared for T639, ECMWF (EC) and Japan models from March to May 2015. The results show that the three models have good performance on predicting the largescale circulation and 850 hPa temperature evolution and adjustment in Asian middle and high latitude areas. EC shows a better performance on westerly index and 850 hPa temperature than the other two models. Taking the sandstorm process from 31 March to 1 April as a case, it is found that all the three models have predicted the intensity and affected range of the major weather for sandstorm, but compared with observations some biases also exist.