Abstract In the last years, the concept of hydraulic flow units (HFU) has been used in the petroleum industry to improve prediction of permeability in uncored interval/wells. This concept is strongly related to the flow zone indicator (FZI) which is a function of the reservoir quality index (RQI). Both measures are based on porosity and permeability of cores. It is assumed that samples with similar FZI values belong to the same HFU. Thus the FZI, along with other significant variables such as porosity, can be used directly to estimate the permeability in a zone; however, the FZI has to be estimated first. In this paper, a novel method based on hybrid soft computing techniques is used for permeability predictions. The technique is known as adaptive network-based fuzzy inference systems or ANFIS, and is based on adaptive neural networks and fuzzy inference systems (FIS). The final inference on FZI is made by a FIS but the parameters of this FIS will be estimated through a learning procedure based on input data; such procedure is typically used in neural network training. The technique is applied in 3 steps: (1) Identification of the dominant variables in rock type behavior, (2) Development of an ANFIS which best suits the real model, using the dominant variables as input and the FZI as output. (3) Estimation of permeability from FZI and porosity values. These steps are applied on a sample case and the results show that hybrid soft computing techniques offer powerful tools for further improving permeability predictions.