Numerical weather prediction (NWP) models, coupled with urban parameterizations, play a crucial role in understanding and forecasting meteorological conditions within urban environments. In the mesoscale NWP model COSMO, only one urban parameterization, TERRA_URB, is available in the model’s operational version. TERRA_URB describes the city as a flat surface with modified physical properties in accordance with the urban canyon geometry. In this study, we have coupled the latest version 6.0 of the COSMO atmospheric model with a more sophisticated urban canopy model, TEB (Town Energy Balance), which explicitly simulates the energy exchange between the facets of the urban canyon. Here, we present the coupling approach and assessment of the model’s sensitivity to urban schemes of different complexity (TEB and TERRA_URB) over the Moscow region for August 2022. Despite using the same external parameters for both schemes, simulations demonstrate notable differences in modeled temperature, with TEB generally producing lower nighttime and morning temperatures. This leads to a greater underestimation of the urban heat island intensity in TEB when compared with the observations but improves the modeled diurnal cycle of the urban temperature. We attribute the observed temperature discrepancies to the different descriptions of heat conductivity and storage within urban surfaces. Although there are no clear advantages to using a more complex parameterization in terms of model air temperature errors, TEB offers more options to fine-tune input parameters and takes into account additional processes, in particular those associated with building heating and cooling, as well as with urban green infrastructure.