Understanding transport diffusion on a molecular level helps to develop improved adsorbents with tailored rate and equilibrium properties. Dual-control-volume grand canonical molecular dynamics (DCV-GCMD) simulations allow the direct simulation of transport diffusion on a molecular level. DCV-GCMD simulations of CH4/CF4 mixtures in carbon nanotubes were carried out. An approach to get composition dependent transport diffusivities directly from DCV-GCMD simulations is introduced. Composition dependent transport diffusivities and fluxes are calculated for varying driving forces in order to investigate the influence of the very large driving force in simulations which is about four orders of magnitude larger than in real experimental systems. Whereas, the flux depends on the driving force the transport diffusivity is independent of it so that DCV-GCMD simulation can be used to simulate transport under experimental conditions. Furthermore, the results of composition dependent diffusivities at four different temperatures are presented. A linear function describes the composition dependence and reproduces the simulated concentration profiles very well. The analysis of the temperature dependence indicates that the transport in the investigated system is due to liquid-like molecular diffusion and not to activated diffusion.