- A number of integral and separate effect experiments were performed in the last two decades for validation of containment computational tools. The main goal of these benchmark experiments was to assess the ability of turbulence models and computational fluid dynamics codes to predict hydrogen concentration distribution and steam condensation rate in a nuclear reactor containment in the course of severe accidents. It appears from the published literature that the predictive capability of the existing computational tools still needs to be improved. This work examines numerically the temporal evolution of helium concentration in the experiment called LOWMA-3, performed in the MISTRA facility of CEA-Saclay, France. In the experiment, helium is used to mimic hydrogen of a real-case accident. The aim of this separate effect experiment, where steam condensation was not involved, is to predict helium concentration field. The conditions of the experiment are such that both the momentum transport and molecular diffusion contributions to the mixing process are of the same order of magnitude ( Fr ∼ 1). A commercial CFD code, Fluent, and a CEA in-house code, Trio_U, are used for flow and helium concentration fields temporal evolution prediction in the present study. The preliminary separate effect studies provide guidance to an optimal modeling approach for the LOWMA-3 experiment. Temporal evolution of helium concentration in the stratification layer is shown, and a comparison to the experiment is discussed. It is shown that correct modeling of the round jet flowfield is essential for a reliable prediction of the mixing process.