The Western Pacific Subtropical High (WPSH) plays an important role in the East Asian climate. The interannual variability of the summer WPSH hindcasted by the GAMIL CliPAS tier-two experiment is investigated. The results show that the Grid Atmospheric Model of State Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics, Institute of Atmospheric Physics, Chinese Academy of Sciences (GAMIL) can reproduce the climatology of the WPSH reasonably. The spatial distribution of predictability index of 500-hPa geopotential height shows that the WPSH has high predictability during 1980-1999. The ensemble mean reasonably reproduces the large variability center of the WPSH, although both the position and the strength of the center are slightly different from the NCEP/NCAR reanalysis results. The interannual variability of the WPSH hindcasted by GAMIL is evaluated by defining a WPSH index according to Sui et al.(2007). The observation evidence shows that the summertime WPSH exhibits significant 2－3 years and 3－5 years oscillations. The strength of 3－5 years oscillation is weaker than that of 2－3 years oscillation during 1980－1999. The GAMIL model captures the 3－5 years oscillation well except with strength stronger than that from the reanalysis. The successful reproduction of 3－5 years oscillation is attributed to the realistic SST in the equatorial central-eastern Pacific used to drive the model. The SST field predicted by Seoul National University (SNU) SST prediction system can well describe the interannual variability of Niño 3.4 SST. For the 2－3 years oscillation of the WPSH, however, the GAMIL hindcast shows poor skill. Further analysis suggests that it is partly due to the weaker variability of SST field at 2－3 years band in the maritime continent (5.5°S－0.5°N, 110.5°E－130.5°E). The quality of predicted SST in the maritime continent partly depends on the quality of historical SST, viz. OISST, which is used in the SNU statistical SST prediction system. The authors suggest that the weaker 2－3 years variability of OISST version 2 in the maritime continent should partly account for the model error.