This paper proposes a novel method using AI algorithms to predict vehicle performance ratings, achieving the goal of predicting vehicle performance ratings from an objective perspective to improve the consistency and accuracy of overall vehicle performance evaluation. Based on field test data, a vehicle dynamics database is established. Using the Pytorch framework, a deep learning intelligent evaluation algorithm is constructed to obtain mathematical relationships between vehicle performance ratings and objective indices, enabling accurate prediction of overall vehicle performance ratings. Finally, field tests verify the effectiveness of the evaluation algorithm. The predicted ratings for handling stability, ride comfort, and steering performance are highly consistent with subjective evaluations by raters, with prediction errors within 0.25 points and maximum relative error of 3.57%. In order to facilitate project application, a web-based human–machine interaction system has been developed to provide technical support for vehicle dynamic performance evaluation and parameter design, as well as optimization and calibration schemes.