Turbulent heat transfer is an extremely complex phenomenon, which has challenged turbulence modellers over various decades. The limitations of the commonly used eddy diffusivity approach have become more evident specially for innovative nuclear reactor applications with low-Prandtl fluids like liquid metals. One of the objectives of the THINS (Thermal Hydraulics of Innovative Nuclear Systems) project sponsored by the European Commission is to push forward the validation and adoption of more accurate closures for single-phase turbulence for innovative reactors in engineering codes. As a part of this THINS project, CD-adapco has implemented in its commercial code STAR-CCM+ an algebraic turbulent heat flux model. In the present work, this implemented model has been widely tested and further calibrated for the application to natural, mixed and forced convection flow regimes at low-Prandtl number. As an outcome, a modelling correlation is proposed in combination with a newly proposed set of model coefficients. This proposed correlation shows dependency of Reynolds and Prandtl numbers in a logarithmic manner to accommodate the wall-normal temperature gradient for the heat flux term. The use of this correlation brings significant improvements in the prediction of heat transfer in liquid metals in all flow regimes. (C) 2014 Elsevier Ltd. All rights reserved.