The Model for Integrated Research on Atmospheric Global Exchanges (MIRAGE) modeling system, designed to study the impacts of anthropogenic aerosols on the global environment, is described. MIRAGE consists of a chemical transport model coupled online with a global climate model. The chemical transport model simulates trace gases, aerosol number, and aerosol chemical component mass (sulfate, methane sulfonic acid (MSA), organic matter, black carbon (BC), sea salt, and mineral dust) for four aerosol modes (Aitken, accumulation, coarse sea salt, and coarse mineral dust) using the modal aerosol dynamics approach. Cloud-phase and interstitial aerosol are predicted separately. The climate model, based on Community Climate Model, Version 2 (CCM2), has physically based treatments of aerosol direct and indirect forcing. Stratiform cloud water and droplet number are simulated using a bulk microphysics parameterization that includes aerosol activation. Aerosol and trace gas species simulated by MIRAGE are presented and evaluated using surface and aircraft measurements. Surface-level SO2 in North American and European source regions is higher than observed. SO2 above the boundary layer is in better agreement with observations, and surface-level SO2 at marine locations is somewhat lower than observed. Comparison with other models suggests insufficient SO2 dry deposition; increasing the deposition velocity improves simulated SO2. Surface-level sulfate in North American and European source regions is in good agreement with observations, although the seasonal cycle in Europe is stronger than observed. Surface-level sulfate at high-latitude and marine locations, and sulfate above the boundary layer, are higher than observed. This is attributed primarily to insufficient wet removal; increasing the wet removal improves simulated sulfate at remote locations and aloft. Because of the high sulfate bias, radiative forcing estimates for anthropogenic sulfur given in 2001 by S. J. Ghan and colleagues are probably too high. Surface-level dimethyl sulfide (DMS) is ~40% higher than observed, and the seasonal cycle shows too much DMS in local winter, partially caused by neglect of oxidation by NO3. Surface-level MSA at marine locations is ~80% higher than observed, also attributed to insufficient wet removal. Surface-level BC is ~50% lower than observed in the United States and ~40% lower than observed globally. Treating BC as initially hydrophobic would lessen this bias. Surface-level organic matter is lower than observed in the United States, similar to BC, but shows no bias in the global comparison. Surface-level sea salt concentrations are ~30% lower than observed, partly caused by low temporal variance of the model's 10 m wind speeds. Submicrometer sea saltis strongly underestimated by the emissions parameterization. Dust concentrations are within a factor of 3 at most sites but tend to be lower than observed, primarily because of neglect of very large particles and underestimation of emissions and vertical transport under high-wind conditions. Accumulation and Aitken mode number concentrations and mean sizes at the surface over ocean, and condensation nuclei concentrations aloft over the Pacific, are in fair agreement with observations. Concentrations over land are generally higher than observations, with mean sizes correspondingly lower than observations, especially at some European locations. Increasing the assumed size of emitted particles produces better agreement at the surface over land, and reducing the particle nucleation rate improves the agreement aloft over land.