Reduced-scale experiments and simulations are important approaches in natural ventilation research, and the similarity requirement is fundamental for generalising the flow characteristics obtained from reduced-to full-scale conditions. However, the similarity requirement of a nonisothermal natural ventilation flow in a reduced-scale model poses additional challenges because of the reduced approaching flow, which can potentially result in Reynolds dependence issues. This study investigated the Reynolds number (Re) independence of indoor airflow in natural ventilation under isothermal and nonisothermal conditions using computational fluid dynamics (CFD) with Reynolds-averaged Navier–Stokes. A wind tunnel experiment was first conducted to validate the accuracy of the CFD using a reduced-scale model. Indoor airflow fields characterised by the same Archimedes number (Ar) but with varying approaching wind velocities and temperatures were compared between the full-scale and 1/10 reduced-scale simulations. The dimensionless ventilation rate showed the least dependence on the Re number, while the temperature field was very sensitive to the Re number, especially in the near-wall region. However, the temperature field on the ventilation pathway is much less dependent on the Re number, the deviation of which is less than 10 % compared to the full-scale simulation. The temperature distribution in the reduced-scale simulation exhibits a thermal stratification pattern similar to that in the full-scale simulation.