A technique for the separation of rarefied gas mixtures based on microelectromechanical systems (MEMS) has recently attracted the attention of researchers. The mechanism of gas separation is the thermally induced flow caused by inhomogeneous temperature gradients within the microchannel, which distinguishes the velocity of different gas species. In this paper, the effects of Knudsen number Kn and equilibrium temperature T0 on the flow properties of each gas species within the gas mixture, such as velocity, molar fraction, streamlines, temperature, and pressure, are investigated by the Direct simulation Monte Carlo (DSMC) method. The flow of the gas mixture is numerically simulated using a quantum scattering-based ab initio (AI) potential and compared with the results of numerical simulations based on the variable soft sphere (VSS) model. The results show that the gas velocity accelerates with increasing temperature, but decreases with increasing Knudsen number. The streamlines are very sensitive to the Knudsen number, yet they are only affected by the temperature in the transitional flow. When the conditions are the same, He flows faster than Ne and the streamlines are smoother. The gas separation factor increases with equilibrium temperature at T0<100K and decreases slightly with equilibrium temperature at T0≥100K. The gas separation efficiency is best in the transitional flow. The temperature discrepancy for the different gas species is less affected by the equilibrium temperature and Knudsen number, with the maximum discrepancy rate being only 2.727% for He and Ne. The pressure discrepancy is more affected by the equilibrium temperature and Knudsen number, with the maximum rate of discrepancy reaching 39.228% for He and Ne. At T0=10K and Kn=1, the discrepancy rates of the gas separation factor, temperature, and pressure obtained based on the AI potential and the VSS model reach maximum values of 44.802%, 4.675%, and 52.716%, respectively. Compared to the VSS model, the AI potential based on quantum scattering is applicable to a wider range of temperatures and is more accurate for both low and high temperatures. Furthermore, the AI potential is more accurate than the VSS model in the transitional flow. The results obtained from the AI potential based on quantum scattering will be more desirable when performing gas separation studies based on thermally induced flows in rarefied gases.