Urban ventilation or wind environment focus on the air flow distribution and pollution dispersion which are studied by the tools of wind tunnel and the CFD. The latter is easy to be employed, but its reliability is under suspicion due to the high uncertainty of air stream. In the city, the sheltering among buildings affects each other in air flow, and the basic element is the two-building model. The paper simulated this essential model with cross-ventilated in a couple of windows on opposite facades and sealed conditions, to compare accurately with the experimental data from the wind tunnel. The analysis of grid sensitivity states that the smallest side length of the cell around 2% of building height is sufficient to obtain steady results. The comparison of five typical RANS model presents: the pressure results in cross-ventilated upstream building are better than that in closed one; estimations on upper parts of building are more accurate than on lower parts; the SST k-ω model can acquire best results with error of 11% in porous interfering building, and with 16% in sealed one, in comparison of experimental data. The methodology of the study could be a guide for the application of research and engineering in urban ventilation.