AbstractImprovement of secondary organic aerosol (SOA) models is critical for accurate understanding of the behavior and sources of atmospheric aerosols. Over the last decade, a number of SOA production pathways were discovered, and several new SOA models have been developed. However, few comparative studies of the performances of the various SOA models have been conducted. In this study, simulation data obtained with five SOA models (two yield models, a volatility basis set (VBS) model, a mechanistic model, and a near‐explicit model) were compared. The performances of the models were evaluated by comparison of the simulated data with observed ratios of the SOA concentration to odd oxygen concentration ([SOA]/[Ox], where [Ox] = [O3] + [NO2]) in the Tokyo metropolitan area. In Tokyo, SOA concentrations have been shown to correlate well with Ox concentrations; thus, Tokyo is an appropriate location for this intercomparison study. All five models showed similar results for the concentrations of gaseous species, including ozone, reactive nitrogen, hydroxy radicals, and volatile organic compounds. In contrast, the simulated SOA concentrations varied substantially among the five models. The VBS model reproduced the observed [SOA]/[Ox] ratio well, whereas the other four models substantially underestimated the ratio. The sensitivity of the ratio to various input parameters differed substantially among the models, as did the volatility distribution of SOA and the source contributions of SOA, suggesting that the choice of SOA model is critical for accurate assessment of the atmospheric behavior and sources of SOA.