An artificial neural network (ANN) model has been developed to generate the multisite streamflows and the results are compared with the classical multsite streamflow generation model developed by Hydrologic Engineering Centre named HEC-4. Both the models have been applied to the case study of Upper Krishna River Basin to evaluate their performances. Important statistical parameters, namely, mean, standard deviation, correlation coefficient of the historical and generated streamflows are compared for the evaluation. Hurst ratio has bem used to' evaluate the strength of persistence of the generated streamflows. This study shows that the streamflows predicted with simple ANN model are more satisfactory than the HEC-4 model in case of multisite streamflow generation.