The stiffness performance of a spatial kinematically redundant parallel mechanism (KRPM) with three relative degree-of-freedom (RDOF) is analyzed. Firstly, the composition principle of the KRPM is introduced, the RDOF is analyzed, the kinematics model is established, and the overall Jacobian matrix is established based on the screw theory. Secondly, the workspace of the KRPM is drawn, and the theoretical stiffness model is established considering the internal restraint force and the external load. The stiffness variation in the workspace is drawn, and the effectiveness of the theoretical stiffness model is verified through finite element analysis. Finally, according to the theoretical stiffness model, four task-based stiffness performance evaluation indexes are obtained, and they are optimized based on the NSGA-II multi-objective genetic algorithm. The results show that adding and adjusting redundant actuators in a parallel mechanism can effectively improve the stiffness performance of the mechanism. In the meantime, different stiffness optimization schemes can be proposed for different task paths based on the NSGA-II genetic algorithm.