A comprehensive dataset on dispersion behind rectangular buildings has been used to analyse the performance of two dispersion models in respect to their handling of building effects: the Danish OML model and the US AERMOD model with the PRIME building algorithm; additionally, the German MISKAM model has been assessed. OML and AERMOD are regulatory plume models with limited requirements in terms of input and computing resources, whereas MISKAM is a computational fluid dynamical model, and as such much more demanding. For most scenarios considered, the degree of misprediction in respect to the maximum concentrations is less than a factor of two for OML and AERMOD. However, in respect to the concentration at a specific location, especially in the near field, both models often result in larger mispredictions. MISKAM provides more accurate predictions.