Increasingly stringent emissions legislation inevitably leads to the requirement of using alternative fuels such as H2. The near field region of an H2 flame is characterized by a high density ratio between the air co-flow and the fuel jet and a high injection velocity, which make turbulence modelling and combustion modelling difficult in this zone. In order to implement H2 efficiently in terrestrial and aerospace power generation applications, the global characteristics of H2 flames need to be well identified. This study presents the effect of turbulence modelling on the predictions of a turbulent non-premixed diluted H2 flame (80 % H2 – 20 % He) characterized by high density and velocity gradients in the inlet region. Two turbulence models, the k-e model and RSM (Reynolds Stress Model), are evaluated in the framework of the SLFM (Steady Laminar Flamelet Model) approach and the results are compared to the experimental data of Barlow and Carter (1994). Mixing is introduced into the predictions analysis using mixture fraction and its variance. These two parameters are included in the transport equations used in the computational fluid dynamics (CFD) code. Computational predictions are compared to measurements at four axial locations. The results demonstrate that the two turbulence models fail to estimate the initial turbulent mixing, which consequently leads to an inaccurate description of the flame characteristics at the first location. Nevertheless, a slight improvement of mixing is achieved using the RSM model. Downstream of the first location, due to the accurate representation of mixing, overall predictions of species and temperature are in good agreement with experimental data. At the second and third location, the prediction accuracy is in favour of the k-e model which describes mixing slightly better; whereas, at the last section, the RSM produces improved results.