The coupled effect of canyon winds and rapid switching of operating scenes deteriorate the aerodynamic performance of the high-speed train (HST) running on the tunnel-bridge section (TBS). What's more, when investigating the aerodynamic performance of the HST by using CFD simulations, the calculation results obtained by various turbulence models show differences, which may have an uncertain effect on analysing the research results. This study aims to analyse the difference in calculation result of three turbulence models, when the HST running on the TBS. For this purpose, based on three mainstream turbulence models (LES, IDDES, and URANS) and the ‘mosaic’ meshing technology, a full-scale CFD dynamic model of 3D train–tunnel–bridge–air was established. Based on the on-site wind field sampling date, UDF program was compiled to load the aforementioned wind-speed time sequence on the velocity-inlet boundary, and the incoming wind field of the TBS was reconstructed. Finally, the aerodynamic load time history and flow field structure of the HST simulated by three turbulence models were revealed, when the HST was running on the TBS. The following conclusions were obtained. The time history of aerodynamic load changes and the fluctuation amplitude calculated by the two turbulence models of IDDES and LES are basically consistent (the difference is kept within 10%), when the HST running on the TBS. The influence of the turbulence model on the aerodynamic loads of the HST is mainly reflected in the head carriage, especially the side force. When the HST runs on the TBS, the side force, lift force, rolling moment, yawing moment and pitching moment of the head carriage calculated by the URANS model are 16%, 9%, 29%, 20%, and 13% higher than the corresponding values of the LES model. Based on the equivalent principle of the side force of the head carriage, a formula is proposed to quantitatively describe the relationship between the side force amplitude of the head carriage on the TBS calculated by the URANS model and the corresponding value of the LES model. The formula can be used to correct the boundary inflow wind speed conditions of the URANS model, so as to quickly obtain the calculation results equivalent to the LES model.