This paper examines the applicability of the following three different combinations of cavitation models to simulate cavitating flows in a nozzle of liquid fuel injector for diesel engines. The first model in a house code consists of the Lagrangian bubble tracking method (BTM), the Rayleigh-Plesset (RP) equation, and large eddy simulation (LES). The second model is the combination of the homogeneous equilibrium model (HEM), a barotropic (Baro) equation, and the RANS turbulence model (k-omega SST). The last one utilizes HEM, RANS (k-epsilon, k-omega SST), and the mass transfer model (MTM), in which bubble dynamics is calculated by the simplified RP equation. OpenFOAM is used for the simulations with the second and third models. Unsteady cavitation in a rectangular injector nozzle is captured by a high-speed camera and the turbulent velocity in the nozzle is measured by laser Doppler velocimetry (LDV); they are compared with the numerical results. As a result, the following conclusions are obtained. The BTM/RP/LES model gives a good prediction for the cavitation length and thickness, as well as cavitation cloud shedding. However, it requires a fine grid and a long CPU time, and is applicable only to incipient cavitation. The HEM/Baro/RANS approach results in a wrong prediction for cavitation length and thickness, and underestimation of the turbulence velocity. It cannot reproduce unsteady cavitation behavior. The combination of HEM/MTM/RANS gives good prediction for the cavitation length and thickness with a relatively coarse grid, and therefore with a short CPU time.