This study addresses the vibration control of magnetorheological (MR) semi-active suspension systems and proposes an optimized model predictive control (MPC) strategy to enhance both ride comfort and driving stability. A seven-degree-of-freedom full-vehicle dynamics model is first developed, and a state-space representation suitable for MPC is constructed. Building upon this model, a real-time MPC algorithm is designed, enabling dynamic adjustment of suspension damping forces under varying operating conditions to achieve superior vibration suppression. Comparative simulations with passive suspension and classical skyhook control demonstrate that the MPC-controlled MR suspension provides substantial improvements in sprung vertical and roll angular accelerations, with marked reductions in vibration indices relative to both benchmarks. Although pitch angular acceleration under MPC is marginally higher than skyhook control in certain scenarios, it consistently outperforms passive suspension. Overall, MPC significantly enhances vertical and pitch ride comfort while maintaining roll stability across different road conditions, thereby exhibiting stronger comprehensive vibration suppression capability. The findings validate the effectiveness and superiority of MPC in MR semi-active suspension systems. The proposed approach not only broadens the application prospects of high-performance suspension technologies but also provides theoretical foundations and practical guidance for the design of intelligent suspension systems.