Interpretation of petrophysical log is one of the most useful and important tools in petroleum geology. Well logs help determine the physical characteristics of a reservoir, such as lithology, porosity, permeability, producer regions and their depth and thickness, and also differentiating between oil and gas and water in a reservoir and defining hydrocarbon reserve. A continual record of the physical characteristics of rocks in various depths is called well logs. Petrophysical logs usually include gamma ray, spontaneous potential, resistance, density, neutron, nuclear magnetic resonance, and sonic. The purpose of this study is to determine those characteristics of the reservoir that cannot be specified directly through present measuring well logs, using an intelligent problem-solving system of neuro-fuzzy. In this study, porosity and permeability are determined as two of the most important reservoir characteristics having much influence on reservoir understanding, reservoir reserve, and capability of reservoir production. The system of determining reservoir characteristics (neuro-fuzzy) was tested on collected well log data from oil and gas fields in the south of Iran. The most important result obtained in this study is that if all data influencing one of the reservoir characteristics are presented to the neuro-fuzzy system, then this system will be an excellent model with low error for determining all of the complex characteristics of the reservoir. These produced intelligent systems predict porosity and permeability completely on training and testing data and the correlation coefficient is near 1 and normalized mean square error is near to zero. Engineers and researchers can predict the reservoir characteristics with very good precision by these intelligent systems. The results of this study prove that a neuro-fuzzy intelligent system is a very powerful tool for determining permeability and porosity at the wells of a naturally fractured gas condensate reservoir. It is the first time to predict porosity and permeability from well logs using adaptive neuro-fuzzy inference system in a naturally fractured gas-condensate reservoir in Khuff Formation (Kangan and Upper Dalan Formation), which is a heterogeneous formation with a wide range of permeability and porosity in the Middle East; this method predicts permeability and porosity precisely in this heterogeneous formation. The other novelty of this study is the choice of appropriate input parameters to determine porosity and permeability precisely. Because of the heterogeneous condition of naturally fractured gas-condensate reservoirs in Khuff Formation, the appropriate input parameters cause the system to train optimally and thus increase the system ability to estimate the output in various conditions precisely.