For a long time, empirical formulars have been used to predict the steady-state creep rate due to lack of clear microscopic description of the mechanism, which frequently leads to unreliable predictions. In this work, a statistical model of single atom developed recently is used to predict the steady-state creep rate at an atomic diffusion level. To test the model, we measure the creep rates of three kinds of materials, i.e., 42CrMoA, 2Cr12Ni, and 1Cr12Mo, and collect the experimental data of other materials, such as IN738LC and K435. The results show that our theoretical predicts are in good agreement with the experimental results.