BACKGROUND Adolescent mental health problems are global public health concern. Primary prevention through physical activity (PA) has been suggested as a potential approach to tackling this problem. Studies in Western countries have provided some evidence of a relationship between PA and adolescent mental health, but the evidence in China is not sufficient. Furthermore, the mechanism behind this relationship has not been empirically tested. The present study aimed at testing the association between PA and mental well-being of Chinese adolescents and to investigate whether a psychological (self-efficacy and resilience) and social (school and family connectedness) mediation model is valid to explain such a relationship. METHODS A total of 775 Chinese students in Grades 7 and 8 were recruited in this cross-sectional study. The participants were given questionnaires to assess their PA level, mental well-being, and the potential mediators. Path models were used to analyse the association between PA and mental well-being, and the roles of potential mediators. RESULTS The PA level was significantly correlated with the adolescent's mental well-being (r = 0.66, p < 0.001), self-efficacy (r = 0.21, p < 0.001), and resilience (r = 0.25, p < 0.001), but not with school connectedness (r = 0.05, p = 0.15) or family connectedness (r = 0.06, p = 0.13). After adjusting for potential confounders in the path model, the PA level was significantly associated with mental well-being (b = 0.52, p < 0.001), and resilience was the only significant mediator (b = 0.31, p < 0.001), which contributed to 60% of this relationship. CONCLUSIONS There was a significant positive association between the PA level and mental well-being of Chinese adolescents. Resilience mediated the majority of this relationship. Promoting physical activities that build up resilience could be a promising way to improve adolescent mental health.