Atmospheric dispersion models are useful tools for assessing air quality and the contributions of emission sources in urban areas. Chemistry and transport models (CTMs) allow evaluation of the role of the regional context, as well as the contribution of the urban background as a whole, but are unable to assess correctly the very local contribution of emission sources located within the city, due to their relatively large grid pitch. For evaluating the contribution of local sources, Lagrangian dispersion models can be used successfully, especially when high spatial resolution modeling is required. In fact, Lagrangian models are based on a more realistic spatialization of urban emissions, which are unevenly distributed due to the layout of the road network (for traffic emissions) and the structure of the built environment (for space heating emissions); In addition, Lagrangian models can also handle building-induced changes in the wind field. In this paper, the simulation results of PM2.5 concentrations in downtown Milan obtained by means of the Lagrangian model AUSTAL2000 are compared with those of the CTM CAMx model. The comparison of simulations referred to an area of 1.7×1.7 km2 in Milan, simulated by both models and focused on three receptor points, selected in order to represent sites with different characteristics in terms of the surrounding environment and different exposure to local emission sources. Specifically, the receptors correspond to a green area, a residential and commercial area near Milan’s main square, and a congested intersection on the inner ring road in the city center. The comparison showed that the results of the Eulerian model on a local scale are representative only of an average background level, similar to the result of the Lagrangian model for the receptor located in a park, but it fails to reproduce concentration gradients at the ‘hot spots,’ which are determined by the highly concentrated emissions produced by local sources, such as road traffic.