Complex turbulent flows such as those encountered in nuclear reactor cooling systems pose considerable challenges for computational fluid dynamics (CFD) simulation using traditional Reynolds-averaged Navier-Stokes (RANS) models based on the linear eddy-viscosity modeling (LEVM) framework. One particular difficulty is the use of low Prandtl number (Pr) fluids such as liquid metal coolants, which considerably alters the fluctuating thermal field and violates the Reynolds analogy upon which turbulent heat flux modeling in LEVMs is based. Although previous studies have shown that Reynolds Stress Models (RSM) offer some improvements over traditional LEVMs for flows containing complex inter-component interaction and Reynolds stress anisotropy, the added complexity, increased computational requirements, and the lack of robustness introduced by traditional RSMs do not always result in an overall improvement. This study evaluates the performance of a newly proposed Algebraic Reynolds Stress Model (ARSM) including an Algebraic Heat Flux Model (AHFM) against two industry standard RANS models, standard k-ε and realizable k-ε model, for a set of canonical test cases relevant to nuclear reactor cooling applications. Numerical simulations using the spectral element code Nek5000 are performed for fully developed channel flows with varying values of Reynolds number (Re) and Pr, both with and without the effects of buoyancy. Results are compared to Direct Numerical Simulation (DNS) data in terms of the velocity and thermal statistics. For all cases investigated, the ARSM model consistently outperforms the other RANS models in this study and it is concluded that the new ARSM model can be a suitable alternative to traditional LEVMs for complex turbulent flows without significant penalty to efficiency and robustness that are commonly associated with traditional RSMs.