In this study, the offline soil moisture simulations from14 land surface models in the Second Global Soil Wetness Project (GSWP-2) is compared and evaluated with the observations from U.S. Department of Agriculture (USDA) Soil Climate Analysis Network (SCAN). The absolute values of soil moisture simulated by models in grid-average still show relative larger bias with observations in point-scale. The different models display distinguished permanent wilt points and field capacities. However, the models show promising ability to simulate the distribution and variability of soil moisture. Removing 10% extreme values, the distribution of simulated soil moisture show good match with the observed values. The “multi-model analysis” displays similar characteristics of inter-annual variance compared with inter-seasonal variance, which indicates the land surface model could capture the major signal of land-atmosphere interaction. The spectrum analysis show great match in the peak frequency of the largest spectrum power.