The objective of this study was to determine the efficiency of the evaluation criteria for comparison some of soil water retention curve models. For this purpose, 70 soil samples were taken from Karaj, Urmia and Shahrekord plains. The soil Water retention curves were obtained using the result of the pressure plate device.Then, the mean of coefficient of determination (R2), the adjusted coefficient of determination (R2 adj) and root mean square error (RMSE) were calculated for each model and the efficiency of these criteria at the evaluation of mentioned models tested with statistical methods. Then, Akaike’s information criterion and Mallows criteria were used for comparison of reference model with other models for each soil. Difference the mean of R2, R2 adj weren’t significant for all soils and textural classes and for RMSE weren’t significant in six classes from eight textural classes (clay and silty clay) at among of the five models. So, R2, R2 adj and RMSE were not suitable criteria for the selection of optimum model (P≤ 0.05). According to Akaike’s information criterion tests Gardner, Pereira and Fredlund, Khlosi and Van Genuchten- Mualem models for 27.1, 10, 45.7 and 68.6% soils and on the basis of Mallows statistic mention models for 34.3, 10, 48.6 and 71.4% soils were better than the Van Genuchten model, respectively. Additionally, Van Genuchten, Van Genuchten- Mualem and Pereira and Fredlund models had the exact prediction from saturation moisture, but Gardner and Khlosi models had larger errors.