Current air pollution levels have been shown to affect human health. Probabilistic modeling can be used to assess exposure distributions in selected target populations. Modeling can and should be used to compare exposures in alternative future scenarios to guide society development. Such models, however, must first be validated using existing data for a past situation. This study applied probabilistic modeling to carbon monoxide (CO) exposures using EXPOLIS-Milan data. In the current work, the model performance was evaluated by comparing modeled exposure distributions to observed ones. Model performance was studied in detail in two dimensions; (i) for different averaging times (1, 8 and 24 h) and (ii) using different detail in defining the microenvironments in the model (two, five and 11 microenvironments). (iii) The number of exposure events leading to exceeding the 8-h guideline was estimated. Population time activity was modeled using a fractions-of-time approach assuming that some time is spent in each microenvironment used in the model. This approach is best suited for averaging times from 24 h upwards. In this study, we tested how this approach affects results when used for shorter averaging times, 1 and 8 h. Models for each averaging time were run with two, five and 11 microenvironments. The two-microenvironment models underestimated the means and standard deviations (SDs) slightly for all averaging times. The five- and 11-microenvironment models matched the means quite well but underestimated SDs in several cases. For 1- and 24-h averaging times the simulated SDs are slightly smaller than the corresponding observed values. The 8-h model matched the observed exposure levels best. The results show that for CO (i) the modeling approach can be applied for averaging times from 8 to 24 h and as a screening model even to an averaging time of 1 h; (ii) the number of microenvironments affects only weakly the results and in the studied cases only exposure levels below the 80th percentile; (iii) this kind of model can be used to estimate the number of high-exposure events related to adverse health effects. By extrapolation beyond the observed data, it was shown that Milanese office workers may experience adverse health effects caused by CO.