The purpose of this study was to examine changes in self-efficacy after adults of a post-stroke exercise program were demonstrated modelling from a peer and non-peer model. We used an ABCA multiple baseline single-subject design with each letter representing a condition: (A) no model/baseline 1 (3-weeks); (B) peer model (6-weeks); (C) non-peer model (6-weeks); and (A) no model/baseline 2 (3-weeks). We recruited participants from Viomax, a Montreal fitness center for persons with physical disabilities. Four participants engaged in the weekly group exercise program for 18 weeks and were presented with a peer model (a fellow person recovering for a stroke) and a non-peer model (a university student) during those respective conditions. Participants completed two self-efficacy questionnaires after each weekly session. Semi-structured interviews were conducted at weeks 9 and 18 of the program. Quantitative visual and trend analysis revealed higher self-efficacy levels for two participants in the peer model and non-peer model conditions when compared to baseline 1. However, self-efficacy ratings appeared to be the highest for the non-peer model condition. Thematic analysis revealed that participants preferred demonstrations from the models as opposed to explanations. Preference for the non-peer model could be because the participants generally had a better relationship with non-peer model. Results provide preliminary indication that modeling, in general, could help people recovering from a stroke increase their self-efficacy, with a slight advantage to non-peer models. Community organizations such as Viomax could implement models in their programs to help increase exercise self-efficacy of their members.