The current interest in hypersonic airbreathing vehicles for new generation launchers and trans-atmospheric vehicles requires a better understanding of physical phenomena within the combustor. In fact, air and fuel must mix and react in a very short time, thus combustion at Mach~ 2œ4 is a critical issue. Numerical simulations can help in understanding how to improve mixing, flame anchoring and combustion efficiency in supersonic reacting flows. Current LES subgrid models developed for subsonic and adapted to supersonic combustion do not predict well or at all experimental results such as flame anchoring whilst past experimental results with hydrogen injected at Mach 2.5 in Mach 2 airstreams showed combustion taking place in about 2 ft. In fact, as shown in previous work, at high Mach number mixing and combustion are driven not only by vortex stretching as in subsonic reacting flows, but also by compressibility and baroclinic effect. Besides, dilatation favours combustion by increasing reaction rates as supersonic combustion occurs locally at about constant volume. All these effects are taken into account in the present (ISCM). Validation of this model has been done by means of two 3-D experimental test cases. The first case is the cross-flow injection of hydrogen (at Mach 1) into an airstream at Mach 2 in a combustor similar to the one built at the University of Tokyo; the second is the supersonic combustion facility built at the NASA Langley Research Centre where injection of hydrogen (at Mach 2.5) is performed at 30 with respect to the airstream at Mach 2. LES simulations by using the known Smagorinsky-Lilly have been alto performed to point out the effect of using different mixing models on the results. Numerical simulations show that the ISCM subgrid model is in better agreement with experimental data than the Smagorinsky-Lilly model: in fact, while the Smagorinsky-Lilly model predicts neither combustion nor vortex structures, ISCM model predicts flame anchoring, streamwise vorticity and temperatures close to those observed in previous experiments.