This paper presents a comparative performance evaluation of the AIRPACT-2 modeling system running with the CALGRID and the CMAQ models separately. The predicted concentrations of ozone (O3), particulate matter (PM2.5), and some selected air toxics species, including formaldehyde, acetaldehyde, 1,3-butadiene, and benzene from both the models have been compared against the observed concentrations. The month of September 2003 was chosen for this study. It was the first month of operation after the original AIRPACT system was updated by incorporating air toxics, and expanding the domain to create AIRPACT-2. The model performances were analyzed in the form of statistical measures including mean bias, fractional bias, absolute mean error, and peak-to- mean ratio. The hourly-observed concentrations of ozone and particulate matter (PM2.5) were collected through a network of surface monitoring stations, and were provided by the Washington Department of Ecology, and the observed concentrations of selected air toxics species were obtained from the EPA Pilot Cities Program. The findings of this study address questions pertaining to the modeling system. An understanding of the factors such as meteorology and emission inventories influencing the results of the AIRPACT-2 system with two different models has also been discussed in this paper.