The chronology of the Inca Empire has traditionally relied on ethnohistoric dates, which suggest that a northern expansion into modern Ecuador began in AD 1463 and a southern expansion into modern Argentina began in AD 1471. We test the validity of these dates with two Bayesian models, which show that the ethnohistoric dates are incorrect and that the southern expansion began before the northern one. The first model of seven dates shows that the site of Chamical, Ecuador, was first occupied cal AD 1410–1480 (95% probability) and has a high probability of being built prior to the ethnohistoric date. The second is an outlier model of 26 14C dates and 19 thermoluminescence (TL) dates from 10 sites along the empire’s southeastern limit in northwestern Mendoza, Argentina. Here, the Inca occupation began cal AD 1350–1440 (95% probability), also earlier than the ethnohistoric date. The model also suggests that the Inca occupation of Mendoza lasted 70–230 yr (95% probability), longer than previously thought, which calls for new perspectives on the timing and nature of Inca conquests and relationships with local groups. Based on these results, we argue it is time to abandon the traditional chronology in favor of Inca chronologies based on Bayesian models.