Background. Predicting walking outcomes poststroke is a challenge for clinicians. Objective. To identify the extent to which exercise dose (repetitions of leg movements) in the first week of a comprehensive stroke unit stay predicts discharge mobility. Methods. A cohort study was conducted on 200 consecutive people admitted to a comprehensive stroke unit who required physical therapy. Results. Discharge and predictor data were available for 191 survivors (99%). On admission, 86 participants were able to walk, and the average walking velocity was 0.42 m/s. On discharge, the average walking velocity was 0.77 m/s, and 152 participants were able to walk. A discharge walking velocity of greater than 0.8 m/s was predicted by the exercise dose achieved in the first week after admission. Adding other predictors did not significantly increase the predictive ability of the model. Completion of more than the median number of exercise repetitions (703) in the first week of admission was associated with a quicker recovery of unassisted walking. This effect persisted after adjustment for walking velocity on admission, cognition, and comorbidity. Conclusion. Exercise dose in the first week after admission for stroke is an important indicator of walking speed at discharge and the time to achieve unassisted walking.