This paper is concerned with model predictive control for polytopic systems subject to exogenous disturbance and H₂/H performance constraints. It is shown that a better characterization of H₂/H performance can be provided by introducing additional free parameters. Meanwhile, input constraints are ensured by using a dilated linear matrix inequality technique, so the utilization of actuator capability is improved. These two aspects bring extra degrees of freedom in optimizing control performance. Recursive feasibility and stability of the controller are proved. Numerical examples verify these properties.