Background Leisure time physical activity (LTPA) is known to be associated with a lower risk for mental health burden, while whether the underlying mechanisms vary across populations is unknown. We aimed to explore the disparate associations between LTPA and mental health based on large-scale data.Methods In this study, we analyzed data including 711,759 individuals aged 15 years or above from the latest four rounds (2003, 2008, 2013, and 2018) of the National Health Service Survey (NHSS) in China. We used multiple logistic regression models adjusted for potential confounders to investigate associations between LTPA and mental health in the total population and subgroups by measuring a diverse set of activity frequencies, intensities, and types. To examine the dose-response associations between total activity volume and mental health, we conducted restricted cubic splines to investigate possible nonlinearity.Results LTPA was associated with remarkably lower self-reported mental health burden (OR 0.56, 95% CI 0.54-0.58). The dose-response relationship between total activity volume and mental health was highly nonlinear (p < 0.001), presenting L-shaped with first 1200 metabolic equivalents of task (METs)-min/week for significant risk reduction (OR 0.58, 95% CI 0.56-0.60). Notably, merely exercising 3-5 times per week with moderate swimming was significantly associated with lower mental health burden among younger people, while the association was strongly large in older adults aged 60 years or above doing 55-min moderate apparatus exercise at least six times a week.Conclusions In a large Chinese sample, LTPA was meaningfully and disparately associated with mental health burden across different people. Policy targeted at prompting activity may be effective for reducing mental health burden, but importantly, tailored strategies are needed based on population contexts.