This study investigates the capability of coupled climate models to predict the interannual variability of the magnitude of vertical zonal wind shear (MWS) in "typhoon season" (from June to October). Seven coupled ocean-atmosphere models used in the DEMETER program have reasonable ability to predict the spatial pattern of the MWS anomalies; however, prediction of false significantly correlated regions between MWS and western North Pacific Typhoon frequency (WNPTF) appears to be their major drawback. MWS has been shown to be positively (negatively) correlated with WNPTF in the low-latitude western North Pacific where significant positive (negative) anomaly of MWS is observed in strong El Niño years, implying that WNPTF in strong El Niño years is more than that in strong La Niña years. The trend is opposite in strong La Niña years. For studying the temporal characteristics, time series of regional average MWS (MWS_T) and time coefficient of the first model form EOF analysis of MWS (MWS_PC1) are defined as MWS indexes. Analyses show that MWS_T and MWS_PC1 are positively correlated with Niño 3.4 index and WNPTF, respectively. A multimodel ensemble (MME) shows significant potential to capture this temporal characteristic. However, the responses of the two MWS indexes to the impact of ENSO are opposite, which cannot be predicted completely by the MME.