We present a probabilistic seismic hazard study for the Caribbean (CAR) that integrates global and regional historic earthquake catalogs, a comprehensive fault database, and geodetic data. To account for the heterogeneity of historic earthquake magnitude types (e.g., , ), we developed regression relationships to convert non‐moment magnitudes to moment magnitudes (). We used a combination of areal sources and fault sources to model seismicity across the entire CAR domain capturing hazard from both shallow and deep earthquakes. Fault sources were modeled using both the characteristic earthquake model of Schwartz and Coppersmith (1984) and the Gutenberg and Richter (1954) exponential magnitude–frequency distribution models, accounting for single and multi‐segment rupture scenarios, as well as balancing of seismic moments constrained by kinematic modeling results. Data from a Global Positioning System survey in conjunction with earthquake information were used to balance seismic moments for different source zones. We also incorporated time‐dependent rupture probabilities for selected faults that have ruptured in recent large earthquakes. The complex tectonics of the CAR and lack of local strong‐motion data necessitates the use of weighted logic trees of the most up to date ground motion prediction equations to account for uncertainty. We present our modeling methodology and hazard results for peak ground acceleration at key return periods, and compare them to recently published regional probabilistic seismic hazard analysis studies.