Mathematical models of vehicle subsystems have main contribution in control and simulation of active safety systems. Among the others, internal combustion engine is a subsystem having high degrees of complexity and nonlinearity, and there is no any accurate and simple model yet being able to predict the engine behavior over the entire range of its variables. In this paper, a semiempirical model is proposed for spark ignition engines that predicts steady torque in terms of throttle position and engine speed. In model development phase, both the physics of the problem and analysis of the measured data are used. Required data for model development are obtained from a validated comprehensive one-dimensional engine model, and the prediction capability of the proposed model is investigated using experimental data. The performance of the model is also compared with a conventional neural networks model. Results show the superiority of the proposed model in comparison with black box models in terms of accuracy, computational cost, and interpretability. This model can be used for determining the required throttle position based on the acceleration demand in longitudinal vehicle dynamics control tasks.