Improving city breathability has been confirmed as one feasible measure to improve pollutant dilution in the urban canopy layer (UCL). Building height variability enhances vertical mixing, but its impacts remain not completely explored. Therefore, both wind tunnel experiments and computational fluid dynamic (CFD) simulations are used to investigate the effect of building height variations (six height standard deviations σH = 0%–77.8%) associated to building packing densities namely λp/λf = 0.25/0.375 (medium-density) and 0.44/0.67 (compact) on city breathability. Two bulk variables (i.e. the in-canopy velocity (UC) and exchange velocity (UE)) are adopted to quantify the horizontal and vertical city breathability respectively, which are normalized by the reference velocity (Uref) in the free flow, typically set at z = 2.5H0 where H0 is the mean building height. Both flow quantities and city breathability experience a flow adjustment process, then reach a balance. The adjustment distance is at least three times longer than four rows documented in previous literature. The medium-density arrays experience much larger UC and UE than the compact ones. UE is found mainly induced by vertical turbulent fluxes, instead of vertical mean flows. In height-variation cases, taller buildings experience larger drag force and city breathability than lower buildings and those in uniform-height cases. For medium-density and compact models with uniform height, the balanced UC/Uref are 0.124 and 0.105 respectively, moreover the balanced UE/Uref are 0.0078 and 0.0065. In contrast, the average UC/Uref in height-variation cases are larger (115.3%–139.5% and 125.7%–141.9% of uniform-height cases) but UE/Uref are smaller (74.4%–79.5% and 61.5%–86.2% of uniform-height cases) for medium-density and compact models.