This work aims to study which level of detail should be preserved in the multibody modelling of a racing car in order to obtain reliable results without excessive model complexity. Three multibody models have been developed and compared through optimal control simulations. The models differ from each other for the order of dynamics comprised: starting from a 14 degrees of freedom (dof) car which includes chassis and wheels motion, a 10 dof model model is obtained by neglecting the wheels hop dynamics, finally a 7 dof model is derived by completely eliminating the suspension motion. Optimal control problem simulations, including parametric analyses varying the center of mass position and suspensions stiffness, have been executed on a full lap on the International circuit of Adria. Simulations results show that the 10 dof model gives almost the same results of the 14 dof one, while significantly reducing the computing time. On the contrary, the basic 7 dof model highlights remarkable differences in both the parametric analysis, suggesting the dynamics has been over-simplified.