Gas flow in coal seam consists of laminar flow through coal cleat and diffusion through pores of coal matrix. Previous studies on the prediction of CBM production mostly focused on the impact of permeability change while the gas exchange between matrix and cleat was assumed to obey unipore diffusion assumption with a single diffusion coefficient. However, numerous scholars have found that a single diffusion coefficient cannot reproduce the sorption kinetic data precisely for a lot of coals, while bidisperse diffusion with fast and slow diffusion coefficients can represent the diffusion process well. Until now, attempts on studying the impact of bidisperse diffusion on CBM production are very limited and mathematical model describing the gas flow with bidisperse diffusion is unavailable. In this study, we propose a fully coupled coal seam gas flow model with consideration of bidisperse diffusion and the interaction between bidisperse diffusion, adsorption strain and geomechanical response of coal. A series of experiments were carried out to understand the characteristics of Sydney Basin required by the gas flow model. The sorption kinetic data was matched by unipore and bidisperse diffusion models, results show that bidisperse diffusion can describe the diffusion process much better than unipore diffusion. Based on the developed gas flow model, the difference of CBM production rates between applying the two diffusion assumptions was studied by using the determined bidisperse diffusion coefficients and the approximated unipore diffusion coefficient. Results show apparent deviations of the predicted production rates, the difference is reduced with decreasing cleat spacing while can still be observed with decreasing initial permeability. From the experimental and modeling results, we believe the assumption of bidisperse diffusion cannot be replaced by unipore diffusion if diffusion is a constraint of gas production. For history matching of field CBM production data, careful examination with consideration of bidisperse diffusion is also recommended to gain a better understanding of in situ permeability change.