Abstract: The vortex initialization scheme used in the operational global model T213L31 for tropical cyclone (TC) track prediction is implemented in higher resolution global spectral model T639L60 in 2009, which is the new generation model for medium range forecast in National Meteorological Center. The vortex initialization scheme used in T213L31 typhoon model includes three parts. One part processes the initial vortex formation using bogus vortex technique when TC being first named in the northwest Pacific Ocean, another part relocates the vortex in the following backgrounds according to the correct position analyzed by the forecasters, and the third part modifies the intensity according to the maximum wind or the minimum sea level pressure analyzed by the forecasters. The comparative analysis of the performance of TC track forecasts for T639L60 and T213L31 typhoon models is carried out using the TC track data from 2009 to 2010 in order to evaluate the ability of TC track prediction of T639L60. The results show that the mean distance errors of T213L31 and T639L61 are close within 72 h forecast, but for 96 h and 120 h forecasts, T639L61 brings a little bit larger errors. The errors of T213L31 and T639L61 for No.0907—No.0922 and No.1001—No.1014 TCs are 120.2 km and 124.8 km (24 h), 211 km and 203.8 km (48 h), 314.7 km and 320.7 km (72 h), 416.3 km and 436.6 km (96 h), 556.3 km and 602.1 km (120 h). The intensity errors of T639L60 are much smaller than those of T213L31 because of the higher resolution of T639L60. The track errors are also analyzed by group all the TCs into five different types of tracks: North-westward moving, northward moving, westward moving, recurvature and unusual track TCs. The results show that T639L60 has better performance for the north-westward moving and recurvature TCs than T213L31, but the track errors are much larger for the northward moving TCs and west moving TCs especially the west moving TCs in the South China Sea. For abnormal track TC forecasts, T639L60 can well forecast the adjustment of the environmental flow and has much smaller mean track errors than T213L31. For the systematic direction errors of landing TC forecasts, T213L31 has the systematic bias of northeast-north and T639L60 has the systematic bias of northeast-east.