Using WRF as a benchmark, GRAMM-SCI simulations are performed for a case study of thermally driven valley- and slope winds in the Inn Valley, Austria. A clear-sky, synoptically undisturbed day was selected when large spatial heterogeneities occur in the components of the surface-energy budget driven by local terrain and land-use characteristics. The models are evaluated mainly against observations from four eddy-covariance stations in the valley. While both models are able to capture the main characteristics of the surface-energy budget and the locally driven wind field, a few overall deficiencies are identified: (i) Since the surface-energy budget is closed in the models, whereas large residuals are observed, the models generally tend to overestimate the sensible and latent heat fluxes. (ii) The partitioning of the available energy into sensible and latent heat fluxes remains relatively constant in the simulations, whereas the observed Bowen ratio decreases continuously throughout the day because of a temporal shift between the maxima in sensible and latent heat fluxes, which is not captured by the models. (iii) The comparison between model results and observations is hampered by differences between the real land use and the vegetation type in the model. Recent modifications of the land-surface scheme in GRAMM-SCI improve the representation of nighttime katabatic winds over forested areas, reducing the modeled wind speeds to more realistic values.