Several criteria have been developed to represent the serviceability conditions of pavement, which international roughness index (IRI) is one of the most important indices. In addition to the application of IRI in case of representing serviceability conditions of the pavement and its application in prioritizing of maintenance and rehabilitation activities, in new mechanistic-empirical methods for design of rigid pavements (e.g MEPDG 2002), it is needed to convert all independent distresses to IRI by utilizing a mathematical model. Thus, determination of IRI on the basis of independent observed distresses is very important. In this research, Adaptive Nero-Fuzzy Inference System (ANFIS) is utilized for modeling of IRI of jointed plane concrete pavements (JPCP) based on long-term pavement performance (LTPP) data. Input data for ANFIS included pavement age in year, initial IRI in m/km, percentage of slabs with transverse cracking, percentage of joints with spalling, pavement surface area with flexible and rigid patching, total joint faulting in mm/km, freezing index in oC and percent subgrade material passing the 0.075-mm sieve and output was considered as IRI in m/km. Results showed that the coefficient of determination (R2) between measured and predicted values of testing set in case of NCHRP equation and ANFIS model are 0.601 and 0.758, respectively. Thus the developed model, based on ANFIS, can be used for accurate predicting of IRI in jointed plane concrete pavements in comparison with NCHRP equation.