The organic Rankine cycle (ORC) represents an emerging technology aimed at exploiting lower temperature heat sources, such as waste heat in industrial processes or exhaust heat in combustion engines. One key aspect of this technology is the efficient and economical operation at part load, typically realized by a partial admission control, which is challenging to predict numerically. Full-annulus computation can only be avoided applying empirical partial admission loss models to conventional full-admission computations. This article aims at assessing the reliability of such a loss model under real-gas and supersonic conditions as a first step toward knowledge-based improved loss models. Three different operating points of an 18.3 kW ORC turbine working with an ethanol–water mixture with two open stator passages (2 × 36 deg) are considered. Full-annulus computational fluid dynamics (CFD) computations are compared to experimental data and results of simulations in a conventional, full-admission, periodic 72 deg-sector model with application of a one-dimensional partial admission loss model. The experimentally obtained mass flow rate and efficiency are matched overall within their measurements accuracy. By highest inlet total pressure, the computed efficiency deviates about 4% from the experiments. Predictions of efficiency based on the full-admission and loss model correction deviate from full-annulus computations less than 1%. These findings suggest that the used empirical correlations for partial admission losses can provide acceptable results in the configuration under investigation.