A model has been developed to predict quantitatively the results of the ASTM D86 fuel distillation procedure. The model uses material and energy balances to treat the procedure as a two stage unsteady-state distillation coupled with an air-filled continuous stirred-tank reactor (CSTR). Heat is removed from the second stage to simulate convection losses from the experimental apparatus. The model requires as inputs the fuel composition and the physical properties of all components (vapor phase heat capacity, vapor pressure, critical properties, density, molecular weight, solubility parameter). Correlations were used to approximate other needed properties. Liquid-phase activity coefficients were calculated with the UNIFAC model. Heat losses were modeled with a correlation from the literature.The model was validated by comparing predictions to experimental measurements on a seven-component model fuel. Agreement was extremely good across the entire range of volume fractions distilled. Agreement was poor during the initial stages if the air-filled CSTR was not included. Compositions available from the model explained that the concentration of each species remained essentially constant until the mixture temperature neared that species' boiling point, after which its concentration fell steadily.D86 curves for model fuels with either ethanol or methyl t-butyl ether (MTBE) present were then compared to predictions from the model. The model correctly predicted the “step” increase in temperature that is observed after ethanol is depleted. The model correctly predicted that MTBE smoothes the D86 curve. The temperature difference between the liquid and vapor regions (measured in a different D86 experiment) was described well by the model for both mixtures. The D86 curve was also predicted for Reddy's synthetic fuel (SAE 861556). Agreement of this model with those data was better than predictions attained from previously reported models.