To model the creep behavior of frozen soils, three creep stages have to be reasonably described(i.e., primary, secondary and tertiary stages). Based on a series of uniaxial creep test results, three creep models were evaluated. It was shown that hypoplastic creep model has high prediction accuracy for both creep strain and strain rate in a wide stress range. The elementary rheological creep model can only be used for creep strains at low stress levels, because of the restraints of its mathematical construction. For the soft soil creep model, the progressive change from the primary to secondary and tertiary stages cannot be captured at high stress levels. Therefore, the elementary rheological and soft soil creep models can only be used for low stress levels without a tertiary stage; while the hypoplastic creep model is applicable at high stress levels with the three creep stages. To model the creep behavior of frozen soils, three creep stages have to be reasonably described (ie, primary, secondary and tertiary stages). Based on a series of uniaxial creep test results, three creep models were evaluated. It was shown that hypoplastic creep model has high prediction accuracy for both creep strain and strain rate in a wide stress range. The elementary rheological creep model can only be used for creep strains at low stress levels, because of the restraints of its mathematical construction. For the soft soil creep model , the progressive change from the primary to secondary and tertiary stages can not be captured at high stress levels. Thus, the elementary rheological and soft soil creep models can only be used for low stress levels without a tertiary stage; while the hypoplastic creep model is applicable at high stress levels with the three creep stages.