A novel gait robot enabled nonambulatory patients the repetitive practice of gait and stair climbing. Thirty nonambulatory patients with subacute stroke were allocated to two groups. During 60 min sessions every workday for 4 weeks, the experimental group received 30 min of robot training and 30 min of physiotherapy and the control group received 60 min of physiotherapy. The primary variable was gait and stair climbing ability (Functional Ambulation Categories [FAC] score 0-5); secondary variables were gait velocity, Rivermead Mobility Index (RMI), and leg strength and tone blindly assessed at onset, intervention end, and follow-up. Both groups were comparable at onset and functionally improved over time. The improvements were significantly larger in the experimental group with respect to the FAC, RMI, velocity, and leg strength during the intervention. The FAC gains (mean +/- standard deviation) were 2.4 +/- 1.2 (experimental group) and 1.2 +/- 1.5 (control group), p = 0.01. At the end of the intervention, seven experimental group patients and one control group patient had reached an FAC score of 5, indicating an ability to climb up and down one flight of stairs. At follow-up, this superior gait ability persisted. In conclusion, the therapy on the novel gait robot resulted in a superior gait and stair climbing ability in nonambulatory patients with subacute stroke; a higher training intensity was the most likely explanation. A large randomized controlled trial should follow.