Abstract: In some literatures, the standard k ε turbulence model has been criticized in predicting flows in strongly curved ducts. However, checking of grid dependency and turbulent parameters were often missing in those papers. In such a situation, a curved duct flow case which had benchmark experimental data was numerically calculated using four meshes. The standard k ε model combined with a wall function was employed for closure of turbulence. Results show that after eliminating the grid dependency, the predicted velocity fields by the standard k ε model can improve agreements with the experiments, compared with the previous calculations. Even in the second half of the bend, the present calculation agrees well with the experimental data. The SST k ω model is also employed for comparison. The numerical results of streamwise and secondary velocities in the second half bend show better agreements with the experiments obtained by the SST k ω model, while for the predicted turbulent fluctuating kinetic energy and stresses, better agreements are shown by the standard k ε model.