This paper presents a numerical study on regenerative cooling in supersonic nozzles, with the aim of identifying the optimal RANS turbulence model for accurately predicting conjugate heat transfer. Three turbulence models, SST, RSM-ω, and RSM-ω with shear flow corrections (SFC), were tested through comparative analysis and simulations to evaluate their accuracy in predicting the heat flux rate and temperature on the nozzle wall. The results indicate that the RSM-ω turbulence model with shear flow corrections provides the best thermal prediction, achieving an improvement of 28% compared with the next best model. The coolant pressure can be very useful in the case of a very hot gas flow, owing to the increase in its saturation temperature. The height of the slot was found to be less important, which means that another factor can be considered a priority when constructing film-cooling systems. The influence of the hydrogen coolant on the gas temperature was also investigated, and it was found that hydrogen was effective in reducing the wall gas temperature, thus preventing the hot sidewall from reaching temperatures that could melt the wall material. The transient results show that the temperature of the side gas increased rapidly as the coolant side after the start-up process, and the time required for cooling the walls was found to be much higher than that of the hot gas flowing in the nozzle. Good agreement was found between the simulation and available experimental data.