Driven with the meteorological data sets based on the reanalyses and gridded observational data archived by the International Satellite Land-Surface Climatology Project (ISLSCP) Initiative II, eleven different land surface models generated global soil moisture data sets for the 10-year period (1986--1995) for the Second Global Soil Wetness Project (GSWP-2). We evaluate these model simulations against in situ observations over grasslands and agricultural regions in the former Soviet Union, United States (Illinois), China, and Mongolia from the Global Soil Moisture Data Bank in terms of their ability to estimate the actual column plant-available soil moisture in the top 1-m soil layer, to simulate the phasing of the annual cycle, and to represent observed interannual variability. Results from these 11 land surface models show that they reproduce reasonably well the observed interannual variability and phasing of the annual cycle. Statistical analysis also shows that the median root mean square of errors among these models ranges from 4 to 8 cm of soil moisture. Similar to what has been found in soil moisture simulations for GSWP-1, the absolute values of soil moisture are poorly simulated by most models. However, the models do a good job of reproducing the soil moisture anomalies. This suggests that the global soil wetness data set produced by GSWP-2 can be used for analyzing climate variability and initializing GCMs by using transform strategies. This also has relevance to subseasonal to seasonal forecasts since soil moisture anomalies may potentially have impact on precipitation.