Models for gas separations with spiral‐wound membranes are developed and found to exhibit good agreement with experiments performed on N2/O2 mixtures. The two‐dimensional (2D) model can be accurately approximated by a one‐dimensional (1D) surrogate model when the spacer widths are chosen to make the channel pressure drops small. Subsequently, the separation of propane/propylene mixtures from the recycle purge stream of a polypropylene reactor is investigated. Assuming ideal gas is found to lead to significant overestimations in membrane stage cuts (sometimes more than 10%), an extent comparable to that associated with extrapolating constant olefin permeance from a low‐pressure condition. While olefin permeance can change significantly with pressure, using a constant‐permeance formulation can result in a small (< 2.5%) underprediction in stage cut if the value for the permeance is taken from the feed condition. Finally, membrane properties and costs necessary for a viable separation process are discussed.