Abstarct: Permeability is an important reservoir property that identify the ability of transition of fluid such as oil, gas and water in rock reservoir. Accurate knowledge about permeability is an effective and suitable tool for petroleum engineers for production and management of oil field. There are two direct and indirect methods for determination of reservoir rock permeability. Direct methods are coring and well testing that both of them take more time and cost. Indirect methods for evaluation of permeability are used well logs by experimental formulas and neural networks as new methods.experimental formulas are so complicated therefor they are not correct. Purpose of this study is use of intelligent systems for determination of permeability, in Sarkhun gas field from available data (well logs and core data). In this work after study of theory principals about permeability, well logs and intelligent systems, horizontal and vertical permeability of Jahrum formation in three wells of Sarkhun gas field were determined by three neural networks called, Early Stopping (ES) back propagation network, Regularization back propagation network and General regression neural network (GRNN), fuzzy system, neuro-fuzzy adaptive network. As results showed, correlation between predict data and core data for two series training and test data in Early Stopping Method are 0.913 , 0.894 , respectively. In regularization Method are 0.979 , 0.886, in General regression Network are 0.91, 0.896, in fuzzy system is 0.919, 0.906 and in neuro-fuzzy adaptive are 0.927 and 0.909.also results show that all methods can predict permeability values in whole of understudy reservoir and neuro-fuzzy adaptive network is better than other methods. For better permeability prediction committee machine with averaging method was used. This method lead to correlation coefficient increase to 0.93 in test step.