•Pressure-based compressible multiphase flow formulation for interfacial and mixture flows.•Flashing in nozzles and valves.•Identified inconsistencies and key assumptions in cavitation model for compressible flow.•Thermodynamic effects in flashing such as variable latent heat was shown to be important for high pressure conditions. A recently published pressure-based compressible multiphase flow model Labois and Narayanan (2017) is validated for applications relevant to safety of pressurized systems, such as flashing of high-pressure water through valves and nozzles. The pressure-based compressible multiphase flow solver is based on non-conservative discretization of the mixture continuity equation and has the advantage of being applicable to interface tracking and n-phase mixture formulations. For simulation of flashing, two well-known cavitation models Singhal et al.(2002), Yuan et al.(2001) have been implemented along with thermodynamic effects such as latent heat of phase change, variable saturation pressure, and heat capacities. The limitations of the well-established cavitation models in terms of their applicability to compressible flows have been clarified. The model was applied to the Super Moby Dick experiment Rousseau (1987). Good flow rates, and pressure and void fraction variations were obtained for three chosen conditions by tuning the cavitation model under incompressible conditions. It was found that the void fraction stops increasing beyond the throat due to thermodynamic effects where the latent heat and heat capacity are functions of pressure and temperature. The reduction of latent heat and rapid changes in liquid and vapour heat capacities close to the saturation line at high pressures was shown to be important. The accuracy, robustness, and efficiency of the pressure-based method has been proven for flashing under strong depressurization conditions. The cavitation model predictions are strongly sensitive to model parameters such as the nucleate density making the models not general enough to be applied to different problems without calibration. It was shown that there is significant room for development of improved cavitation models especially focussing on the constant number density constraint due to the strong sensitivity of the current models to this input parameter. The assumption of incompressibility and homogeneity of the phasic velocities is also identified as serious limitations requiring further study.