This study presents the performance analysis of multi- segment continuum robots. Since continuum robots are designed to provide excellent dexterity, two local indices, axiality and angularity dexterity, are introduced to study the dexterity that is inspired by separating Jacobian matrix. A Monte Carlo Method is adopted to simulate the distribution of local dexterity over the workspace. On this basis, the corresponding global indices in axiality and angularity are defined to compare global dexterity performance. To investigate the optimal kinematic performance, an objective function related to the segment lengths is designed under the consideration of reachable workspace as well as dexterity performance. Particle Swarm Optimization (PSO) algorithm is adopted to solve the optimization problem successfully. The optimal length distributions for two-segment and three-segment continuum robots are discovered. It is found that this method can also apply to general multi-segment continuum robots.