Accurate subgrid-scale (SGS) scalar flux models are essential when large eddy simulation (LES) is used to represent flow, mixing and transport of passive and active scalars in engineering, and environmental applications in turbulent regime. Many SGS scalar flux models have been developed for flows with low Schmidt numbers (Sc), but their application to high Sc flows has important limitations. In high Sc flows, the behavior of the scalar field becomes anisotropic because of intermittency effects, phenomenon that must be accounted for by SGS scalar flux models. The objective of this paper is to evaluate the ability of three SGS scalar flux models to predict the scalar behavior of a high Sc-number flow configuration, namely the anisotropy-resolved SGS scalar flux model: (1) appropriate for high Sc-number flow configurations, and two additional SGS models (linear eddy diffusivity based SGS models) with (2) constant, and (3) dynamically calculated turbulent Schmidt number. The LES simulation results accomplished by these models are compared to each other and to experimental data of a turbulent round jet discharging a diluted scalar into a low-velocity coflowing water stream. The comparison of simulation results and experimental observations shows that, in general, all SGS models reproduce the mean filtered concentration distribution in radial direction. The dynamic eddy diffusivity and anisotropy models reproduce the rms of the concentration and SGS scalar fluxes distribution. In particular, the anisotropy model improves the prediction reliability of LES. However, the three models evaluated in this study cannot accurately predict the scalar behavior at the superviscous layer. Finally, this work demonstrates that complex models can achieve reliable predictions on reasonable grids using less computational effort, while simple models require fine grids with increased computational costs.