A quasi-dimensional (QD) simulation model is a preferred method to predict combustion in the gasoline engines with reliable results and shorter calculation time compared with multi-dimensional simulation. The combustion phenomena in spark ignition (SI) engines are highly turbulent, and at initial stage of the combustion process, turbulent flame speed highly depends on laminar burning velocity SL. A major parameter of the QD combustion model is an accurate prediction of the SL, which is unstable under low engine speed and ultra-lean mixture. This work investigates the applicability of the combustion model for evaluating the combustion characteristics of a hightumble port gasoline engine operated under ultra-lean mixture (equivalence ratio up to φ=0.5) which is out of the range of currently available SL functions initially developed for a single component fuel. In this study, the SL correlation is improved for a gasoline surrogate fuel (5 components). Predicted SL data from the conventional and improved functions are compared with experimental SL data taken from a constant-volume chamber under micro-gravity condition. The SL measurements are done at reference conditions at temperature of 300K, pressure of 0.1MPaa, and at elevated conditions whose temperature = 360K, pressure = 0.1, 0.3, and 0.5 MPaa. Results show that the conventional SL model over-predicts flame speeds under all conditions. Moreover, the model predicts negative SL at very lean (φ ≤0.3) and rich (φ ≥1.9) mixture while the revised SL is well validated with the measured data. The improved SL formula is then incorporated into the QD combustion model by a userdefined function in GT-Power simulation. The engine experimental data are taken at 1000 RPM and 2000 RPM under engine load IMEPn=0.4-0.8 MPa (with 0.1 increment) and φ ranges are up to 0.5. The results shows that the simulated engine performances and combustion characteristics are well validated with the experiments within 6% accuracy by using the QD combustion model coupled with the improved SL. A sensitivity analysis of the model is also in good agreement with the experiments under cyclic variation (averaged cycle, high IMEP or stable cycle, and low IMEP or unstable cycle).