Background: Although stroke survivors can benefit from robotic gait rehabilitation, stationary robot-assisted gait training needs further investigation. In this paper, we investigated the efficacy of this approach (with an exoskeleton or an end-effector robot) in comparison to the conventional overground gait training in subacute stroke survivors. Methods: In a multicenter controlled clinical trial, 89 subacute stroke survivors conducted twenty sessions of robot-assisted gait training (Robotic Group) or overground gait training (Control Group) in addition to the standard daily therapy. The robotic training was performed with an exoskeleton (RobotEXO-group) or an end-effector (RobotEND-group). Clinical outcomes were assessed before (T0) and after (T1) the treatment. The walking speed during the 10-Meter Walk Test (10 MWT) was the primary outcome of this study, and secondary outcomes were the 6-Minute Walk Test (6 MWT), Timed Up and Go test (TUG), and the modified Barthel Index (mBI). Results: The main characteristics assessed in the Robotic and Control groups did not differ at baseline. A significant benefit was detected from the 10 MWT in the Robotic Group at the end of the study period (primary endpoint). A benefit was also observed from the following parameters: 6 MWT, TUG, and mBI. Moreover, patients belonging to the Robot Group outperformed the Control Group in gait speed, endurance, balance, and ADL. The RobotEND-group improved their walking speed more than the RobotEXO-group. Conclusion: The stationary robot-assisted training improved walking ability better than the conventional training in subacute stroke survivors. These results suggest that people with subacute stroke may benefit from Robot-Assisted training in potentiating gait speed and endurance. Our results also support that end-effector robots would be superior to exoskeleton robots for improving gait speed enhancement.