We report kinetic Monte Carlo (KMC) simulations of self-diffusion of a methane/perfluoromethane mixture in silicalite. The hopping rates and model setup were taken to match previously published MD simulations and PFG-NMR data: In the case of the present KMC simulation the acceleration/deceleration of particles with varying mixture composition can be attributed to correlation effects. The logarithmic interpolation rule for mixture self-diffusion coefficients reported by Snurr and Karger (J. Phys. Chem. B 1997, 101; 6469) is validated for a broad range of loadings. However, a deficiency of the present KMC model is that it is not able to cope with different saturation capacities of the two species; the influence of different saturation capacities on mixture diffusion is accounted for by the Maxwell-Stefan theory. The Maxwell-Stefan formulation of diffusion in multicomponent mixtures is used to obtain explicit formulas for calculating the diffusivities of binary mixtures within a zeolite matrix. The theoretical development allows the estimation of the mixture diffusivities on the basis of the pure component diffusivities at zero loadings. We discuss several mixture rules for the estimation of the exchange coefficient D-12. Though none of them is fully satisfying, we demonstrate the D-12 value should have the same order of magnitude as the pure component Maxwell-Stefan diffusivities in order to account properly for acceleration/deceleration behavior.