This cross-sectional study examined self-efficacy as a predictor of physical activity (PA) levels among 484 adolescents in Toronto. Participants completed four PA measures (three single-items; frequency and duration of PA listed, done in the past seven days), a 24-item measure of self-efficacy to overcome barriers to PA, and a demographics measure. Factor analysis of the self-efficacy measure yielded five subscales: internal, personal safety, physical environment, social environment, and responsibilities barriers. The subscales had high internal consistency reliability. Regression analyses, when controlling for sex, age, and BMI, indicated that (a) self-efficacy to overcome internal barriers significantly predicted moderate-to-vigorous, vigorous, and moderate PA (9%, 6%, and 8% additional variance explained, respectively) and (b) self-efficacy to overcome social and physical environment barriers significantly predicted MET hours per week of PA (13% additional variance explained). The results can be used to identify areas to target for developing interventions to increase physical activity levels.