Emissions from major point sources are badly represented by classical Eulerian models. An overestimation of the horizontal plume dilution, a bad representation of the vertical diffusion as well as an incorrect estimate of the chemical reaction rates are the main limitations of such models in the vicinity of major point sources. The plume-in-grid method is a multiscale modeling technique that couples a local-scale Gaussian puff model with an Eulerian model in order to better represent these emissions. We present the plume-in-grid model developed in the air quality modeling system Polyphemus, with full gaseous chemistry. The model is evaluated on the metropolitan Île-de-France region, during six months (summer 2001). The subgrid-scale treatment is used for 89 major point sources, a selection based on the emission rates of NO_x and SO₂. Results with and without the subgrid treatment of point emissions are compared, and their performance by comparison to the observations on measurement stations is assessed. A sensitivity study is also carried out, on several local-scale parameters as well as on the vertical diffusion within the urban area. <br><br> Primary pollutants are shown to be the most impacted by the plume-in-grid treatment. SO₂ is the most impacted pollutant, since the point sources account for an important part of the total SO₂ emissions, whereas NO_x emissions are mostly due to traffic. The spatial impact of the subgrid treatment is localized in the vicinity of the sources, especially for reactive species (NO_x and O₃). Ozone is mostly sensitive to the time step between two puff emissions which influences the in-plume chemical reactions, whereas the almost-passive species SO₂ is more sensitive to the injection time, which determines the duration of the subgrid-scale treatment. <br><br> Future developments include an extension to handle aerosol chemistry, and an application to the modeling of line sources in order to use the subgrid treatment with road emissions. The latter is expected to lead to more striking results, due to the importance of traffic emissions for the pollutants of interest.