This study provides a comprehensive evaluation of historical surface soil moisture simulation (1979–2012) over Eurasia at annual and seasonal time scales between two medium-resolution versions of the Beijing Climate Center Climate System Model (BCC-CSM)—one that is currently participating in phase 6 of the Coupled Model Intercomparison Project (CMIP6), i.e., BCC-CSM2-MR, and the other, BCC-CSM1.1m, which participated in CMIP5. We show that BCC-CSM2-MR is more skillful in reproducing the climate mean states and standard deviations of soil moisture, with pattern correlations increased and biases reduced significantly. BCC-CSM2-MR performs better in capturing the first two primary patterns of soil moisture anomalies, where the period of the corresponding time series is closer to that of reference data. Comparisons show that BCC-CSM2-MR performs at a high level among multiple models of CMIP6 in terms of centered pattern correlation and “amplitude of variation” (relative standard deviation). In general, the centered pattern correlation of BCC-CSM2-MR, ranging from 0.61 to 0.87, is higher than the multi-model mean of CMIP6, and the relative standard deviation is 0.75, which surmounts the overestimations in most of the CMIP6 models. Due to the vital role played by precipitation in land–atmosphere interaction, possible causes of the improvement of soil moisture simulation are further related to precipitation in BCC-CSM2-MR. The results indicate that a better description of the relationship between soil moisture and precipitation and a better reproduction of the climate mean precipitation by the model may result in the improved performance of soil moisture simulation.