In the present paper, the analysis of concentration and flow fields around a model building was performed using two different approaches in turbulence modeling. In the first approach, the non-linear model of Ehrhard and Moussiopoulos was employed as one of the best cubic non-linear eddy viscosity models, and large eddy simulation was utilized in the second approach. The obtained results suggest that although the non-linear model has the ability to predict the anisotropic normal components of Reynolds stress tensor, due to time-averaged nature of its governing equations, it is incapable of estimating the correct values of Reynolds stress components and turbulent characteristics in the wake region of the model building. So the predicted turbulence kinetic energy values by the non-linear model are 45% smaller than those of large eddy simulation approach in the wake region behind the model building. Also different predictions of the wake region structure by the non-linear model and large eddy simulation approach revealed that elongation of iso-surface for \(\left\langle K \right\rangle\) = 1.5 in non-linear model is 56% more than that of large eddy simulation approach. The large eddy simulation approach shows much consistent behavior to the physics governing the flow compared to the non-linear model. Also good agreement observed between the results obtained through this approach and the experimental data. However, the disadvantage of large eddy simulation approach is the high computational costs.