Fine particulate matter (PM) has been linked to gestational diabetes mellitus (GDM). However, PM is a complex mixture with large spatiotemporal heterogeneities, and women with early-onset GDM (i.e., diagnosed before 24 gestation week) have distinct maternal characteristics and a higher risk of worse health outcomes compared with those with late-onset GDM (i.e., diagnosed in or after 24 gestation week). We aimed to examine differential impacts of PM and its constituents on early- vs. late-onset GDM, and to identify corresponding susceptible exposure windows. We leveraged statewide linked electronic health records and birth records data in Florida in 2012-2017. Exposures to PM and its constituents (i.e., sulfate [SO ], ammonium [NH ], nitrate [NO ], organic matter [OM], black carbon [BC], mineral dust [DUST], and sea-salt [SS]) were spatiotemporally linked to pregnant women based on their residential histories. Cox proportional hazards models and multinomial logistic regression were used to examine the associations of PM and its constituents with GDM and its onsets. Distributed non-linear lag models were implemented to identify susceptible exposure windows. Exposures to PM, SO , NH , and BC were statistically significantly associated with higher hazards of GDM. Exposures to PM during weeks 1-12 of gestation were positively associated with GDM. Associations of early-onset GDM with PM in the 1 and 2 trimesters, SO in the 1 and 2 trimesters, and NO in the preconception and 1 trimester were considerably stronger than observations for late-onset GDM. Our findings suggest there are differential associations of PM and its constituents with early- vs. late-onset GDM, with different susceptible exposure windows. This study helps better understand the impacts of air pollution on GDM accounting for its physiological heterogeneity.