Mixing is a key point for thrust and efficiency of combustion systems. It becomes crucial in the case Liquid Rocket Engines as large investments are involved. Besides, the pressure in liquid rocket combustion chamber often exceeding the critical point of loaded propellants, mixing becomes an important scientific issue as fluid properties differ from classical ideal gas assumption. In this study, two configurations are studied to evaluate the impact of subgrid models on mixing. Firstly, Mayer's experiments of trans- and supercritical nitrogen jet injection into a warm nitrogen atmosphere have been numerically investigated with a structured numerical code called SiTCom-B. SiTCom-B solves Direct Numerical Simulations and Large Eddy Simulations equations for perfect or real gas equation of states. In this study, Soave-Redlich-Kwong (SRK) and Peng-Robinson equation of state are used with appropriated thermodynamics relations and validated against NIST data. Three-dimensional LES are conducted for two cases (cases 3 and 4 in Mayer et al.'s reference [1]) with real-gas NSCBC treatment. Several sub-grid scale models are tested and the results are compared to experimental data for density on jet axis: a very good agreement is obtained on a light mesh (11.6 million of points) with the SRK equation of state and standard Smagorinsky model. Flow structures are evidenced with Schlieren snapshots. Secondly, the Mascotte test-bench from ONERA is simulated with SiTCom-B based on Soave-Redlich-Kwong equation of state and Smagorinsky models. The simulated non-reacting case is characterized by a very short liquid oxygen dense core because of the strong density and velocity gradients boosting mixing efficiency. (C) 2013 Elsevier B.V. All rights reserved.