There is a need to compare and understand the characteristics of the land surface models used in the GAME (GEWEX Asian Monsoon Experiment) project with a common forcing data in order to integrate the various products of the project. We successfully drove the JMA-SiB, CCSR/NIES-LAND, MATSIRO and TOHOKU models, using the GAME/Tibet POP97 data (during about one month, with a time interval of one hour) at the Tanggula river basin (D105). This dataset also has some validation data (soil temperature and snow depth). All models succeeded in forecasting the soil freezing process, since there were no large differences in the ground surface flux. However, there were large differences in the Bowen ratio due to the differences in parametarization of stomatal resistance. It is expected that these differences in the Bowen ratio will cause a large difference when these land surface schemes are coupled with GCM or reg-CM. Large differences in snow amount are attributable to the differences in the estimation of albedo.