Several criteria exist for determining the optimal design for a surgical robot. This paper considers kinematic performance metrics, which reward good kinematic performance, and dynamic performance metrics, which penalize poor dynamic performance. Kinematic and dynamic metrics are considered independently, and then combined to produce hybrid metrics. For each metric, the optimal design is the one that maximizes the performance metric over a specific design space. In the case of a 2-DOF spherical mechanism for a surgical robot, the optimal design determined by kinematic metrics is a robot arm with link angles (α₁₂=90°, α₂₃=90°). The large link angles are the most dextrous, but have the greatest risk of robot-robot or robot-patient collisions and require the largest actuators. The link lengths determined by the dynamic metrics are much shorter, which reduces the risk of collisions, but tend to place the robot in singularities much more frequently. When the hybrid metrics are used, and a restriction that the arm must be able to reach a human's entire abdomen, the optimal design is around (α₁₂=51°, α₂₃=54°). The hybrid design provides a compromise between dexterity and compactness.