This article reports the importance of the correct prediction of turbulent heat transfer in liquid metal flows with the use of Reynolds-Averaged Navier-Stokes (RANS) modelling approach. The Prandtl number of liquid metals is of the order of 0.025-0.001, hence making it very challenging for “off-the-shelf” RANS models to correctly predict the heat transfer phenomena. In this regard, some peculiarities of the “off-theshelf” RANS models are highlighted in different flow regimes, i.e. natural, mixed and forced convection. Furthermore, an updated status and perspectives of the available turbulent heat flux modelling closures within the nuclear community are provided with an especial focus on the low-Prandtl fluids.