Liquid metal–cooled reactors are expected to play an important role in the future nuclear energy landscape, thanks in part to their higher fuel utilization rates compared to conventional water-cooled reactors. These reactors are typically cooled using sodium or lead (alloys), which have low Prandtl numbers, making the modeling of heat transfer in these reactors challenging. The use of the Simple Gradient Diffusion Hypothesis (SGDH) for the calculation of the turbulent heat flux has become a standard computational fluid dynamics practice because of the reasonable results obtained with this approach for many fluids. However, the SGDH is known to give inaccurate results for low-Prandtl-number fluids. In order to overcome this limitation of the SGDH, an Algebraic Heat Flux Model (AHFM) was developed in recent years that can improve the accuracy of heat flux modeling in low-Prandtl-number fluids. In previous works, this model’s coefficients were computed from global flow parameters such as the Reynolds number, which may not necessarily be known in advance and may even be impossible to define in transient analyses or complex flow configurations. In this work, we propose an alternative formulation of the AHFM coefficients, in which they are calculated automatically from local turbulence parameters. We validate this new formulation with direct numerical simulation results in several different forced convection flow configurations and demonstrate that it provides an improvement in the prediction of the mean temperature field compared to the SGDH. Additionally, we show that the new formulation, despite the automatization of the coefficients, performs at least as well as the original AHFM formulation for the selected cases at low Prandtl numbers.