Direct-injection spark-ignition (DISI) engines are regarded as a promising technology for the reduction of fuel consumption and improvement of engine thermal efficiency. However, due to direct injection, the shortened fuel-air mixing duration leads to a spatial gradient of the equivalence ratio, and these locally rich regions cause the formation of particulate matter.In the current study, numerical investigations on pollutant formation in a DISI engine were performed using combined flamelet models for premixed and diffusion flames. The G-equation model for partially premixed combustion was improved by incorporating the laminar flamelet library. Gasoline fuel was represented as a ternary mixture of gasoline surrogate and its laminar flame speeds were obtained under a wide range of engine operating conditions. For the flame propagation in a partially premixed condition, the presumed shape of the probability density function approach was adopted, whereas the burned gas compositions were determined from the steady laminar flamelet library. Aided by the detailed composition of the burned gas, this model is able to predict the major pollutants, i.e., carbon-monoxide and nitric-oxide. The phenomenological model was implemented to simulate the particulate matter formation.The numerical analyses were conducted under variation in the engine operating conditions while varying the engine speed, intake pressure and ignition timing. The simulation results were well matched with the experimental results from the single-cylinder DISI engine. The oxidation of carbon monoxide was captured near the stoichiometric region as a development of the secondary diffusion flame, whereas the NO emission levels are in good agreement with the experimental data while the spatial distributions are in accordance with the mixture and temperature fields.