Wave characteristics are one of the most important factors in design of coastal and marine structures. Therefore, an accurate prediction of wave parameters is considerably important. In this paper, SWAN and MIKE 21 SW third generation spectral models have been used for the prediction of wave parameters. The field data set of Lake Erie has been used for testing the performance of the models. Significant wave height (Hs), peak spectral period (Tp) and mean wave direction were hindcasted in the study. Both models were forced by temporally varying wind. The results show that the average scatter index of SWAN is about 16% for Hs and 19% for Tp; while the average scatter index of MIKE 21 SW is about 20% and 13% for Hs and Tp, respectively. The inconsistency between the results of the models was found to be due to differences between the wind input parameterizations. Using Komen’s formulation for the wind input led to a more accurate prediction of Hs rather than using Janssen’s formulation for the wind input. It was also found that using the cumulative steepness method for whitecapping dissipation in SWAN model yields a less accurate estimation of Hs and a more accurate estimation of Tp. By using this method, the average scatter index increased about 7% for Hs prediction and decreased more than 6% for Tp prediction. In addition, the computational time required for cumulative steepness method is more than 2 times of Komen’s option. Both models were also evaluated for the prediction of wave direction and it was found that MIKE 21 SW results are slightly more accurate than those of SWAN.