Steam condensation with non-condensable gas (air, helium, etc.) is a vital heat and mass transfer phenomenon observed in the field of nuclear energy, different from pure steam condensation, with steam condensing, non-condensable gas will precipitate on the condensation surface and attach to form non-condensable gas layer with high thermal resistance, which will seriously inhibit the condensation heat transfer. Initially, research on this topic primarily relied on experimental methods; however, with the advancement of CFD, numerical simulation has gained significant attention. The initial CFD for steam–air condensation were predominantly based on experimental correlation models to determine the condensation rate. Although these calculations are efficient, they are limited by the range of experimental parameters and fail to capture the heat and mass transfer process near the wall. In this paper, a condensation model is established based on the diffusion boundary layer theory, which enables simulation of local condensation processes, and the suitability of different turbulence models and wall functions is analyzed using COPIAN experiment. The chosen turbulence model is the commonly employed k-ε and k-ω model, while the corresponding wall treatment methods mainly involve low-y+ and high-y+. It is found that the k-ε model exhibits smaller calculation error compared to experimental data. The maximum average heat flux calculation error of the k-ω model is 12.47%. For various wall treatment functions, it has been observed that the calculation error for the low-y+ approach in the k-ε model can exceed 50% but smaller in k-ω model, which means the k-ε turbulence model all-y+ or two layer all-y+ approach is advised.