Neuro-fuzzy (NF) systems were applied to develop a new model for enhancing the prediction accuracy of breach formation time of embankment dams (t(f)), which is widely recognized to have uncertainty and affects the accuracy of dam breaks simulations. NF is based on expert knowledge which is trained by a learning algorithm derived from neural network theory and defined by a set of IF-THEN rules, each of which establishes a local linear input-output relationship between the variables of the model. The erodibility of the embankment material, the height of water and volume of water above the breach invert were taken as the input variables that contribute to t(f). Historical data from 45 embankment dam failures were randomly divided into two sets and used to train and test the performance of the developed model. Two models, without and with inclusion of the erodibility as input variable, were addressed. To train the models two different output types with ten different combinations of input fuzzy sets were examined. Therefore, 40 different NF models were investigated in order to obtain the best model that adequately resembles the observed t(f) values of the testing dataset. The results of the NF model were also compared with the results of the best available regression models (RM) and a recent gene expression programming (GEP). Three statistical evaluation parameters were used to evaluate the results of the models. Test results show the potential of the developed NF model which performed better than the available RM and GEP.