Laminarization of a turbulent flow due to wall heating has been known for more than 50 years, to the point that it is sometimes used as means of reducing friction. However this phenomenon has been mainly studied for cylindrical pipes and with imposed heat flux but not for channel flows and with imposed temperature boundary conditions, especially with asymmetric ones (that is to say in presence of a transverse thermal gradient). Based on the recent success of some Reynolds-averaged Navier-Stokes (RANS) models to correctly describe the influence of a strong transverse temperature gradient on turbulent Poiseuille flows, when compared to similar direct numerical simulations (DNS) or large eddy simulations (LES) results, these approaches are used here to investigate reverse transition. Since the choice of turbulence model has a non-negligible influence on the results, however, it is necessary to use different models to get an indication of the uncertainty associated with them. The proposed methodology is based on the use of RANS closures that do not involve any wall functions due to the strong gradient in the wall layer that has to be modeled. Thus, two first-moment closures and a second-moment closure are considered: the k − ω − SST and the k − ε − v 2 /k, and the EB-RSM. The latter two rely on an elliptic blending. The turbulent heat flux is modeled with a simple gradient diffusion hypothesis (SGDH) and a generalized gradient diffusion hypothesis (GGDH) for the first-moment and second-moment closures respectively. In summary, more than 800 calculations are performed for the above three models in order to analyze the reverse transition, and to open room for debate on the possibility for such approaches to correctly reproduce the experimentally observed behavior.