The need of exchange rate forecasting in order to preventing its disruptive movements has engrossed many policy makers and economists for many years. The determinants of exchange rate have grown manifold making its behavior complex, nonlinear and volatile so that nonlinear models have better performance for its forecasting. Nonlinear models estimated by various methods can fit a data base much better than linear models. Beside they can learn from examples, are fault tolerant in the sense that they are able to handle noisy and incomplete data, are able to deal with non-linear problems and once trained can perform prediction and generalization at high speed. In this study, the accuracy of ANFIS and ANN as the nonlinear models and GARCH and ARIMA as the linear models for forecasting 2, 4 and 8 days ahead of daily Iran Rial/ and Rial/US$ was compared. Using three forecast evaluation criteria (R2, MAD and RMSE) we found that nonlinear models outperform linear models, GARCH outperforms ARIMA model and ANFIS outperforms ANN model. And consequently the effective role of ANFIS model to improve the Iran’s exchange rate forecasting accuracy can’t be denied.