The very large number of distinct structures that are known for metal-organic frameworks (MOFs) and related materials presents both an opportunity and a challenge for identifying materials with useful properties for targeted applications. We show that efficient computational models can be used to evaluate large numbers of MOFs for kinetic separations of light gases based on finding materials with large differences between the diffusion coefficients of adsorbed gas species. We introduce a geometric approach that rapidly identifies the key features of a pore structure that control molecular diffusion and couple this with efficient molecular modeling calculations that predict the Henry's constant and diffusion activation energy for a range of spherical adsorbates. We demonstrate our approach for >500 MOFs and >160 silica zeolites. Our results indicate that many large pore MOFs will be of limited interest for separations based on kinetic effects, but we identify a significant number of materials that are predicted to have extraordinary properties for separation of gases such as CO2, CH4, and H2.