In an earlier research (Chen et al., 2015a) a mathematical model was established to simulate tracer mixing (a KCl solution). The predicted Residence Time Distribution (RTD) curves showed good agreements with experimental RTD curves for larger amounts of tracer additions. However, for smaller additions (50mL) of a KCl solution into water, the predicted RTD curves tended to deviate from the experimental RTD curves for a tundish (a continuous flow reactor). The current paper focuses on the possibilities that the predictability for smaller additions could be resolved by using a suitable turbulence model. The performance of five different turbulence models representing different modeling techniques and levels of complexity were tested in combination with using a density-coupled mixed composition fluid model to simulate the mixing, i.e. the following models: LVEL, Chen–Kim k–ε, MMK k–ε, Explicit Algebraic Reynolds Stress Model (EARSM), and Large Eddy Simulation (LES): Wall-Adapting Local Eddy-viscosity (WALE). The results indicate that models that tend to resolve turbulence structures renders better predictions of the mixing process of smaller tracer amounts. In addition, the influence of different tracer amounts on the flow in tundish was assessed. The simulation results for 75mL, 100mL, 150mL, and 250mL KCl tracer additions were compared. The results showed that in an upward flow the tracer will, sooner or later (dependent on the tracer amount), sink to the bottom. This is due to the higher density of the tracer compared to the density of water. From a physical modeling perspective, this issue is like the “butterfly effect”. It is showed that for a slight increase of the amount of tracer, the flow field might be disturbed. This, in turn, will result in a shifted RTD curve.