The timing, duration and evolution of sea level during the Marine Isotope Stage 5e (MIS 5e) highstand is a subject of intense debate. A major problem in resolving this debate is the difficulty of chronologically constraining the sea level fall that followed the peak of the highstand. This was mainly controlled by ice-sheet dynamics, the understanding of which is relevant for assessing future sea-level behavior due to global warming. Here we use stratigraphical and geochoronological (U/Th dating and tephra fingerprinting) evidence from the Infreschi archaeological cave (Marina di Camerota, Southern Italy) to constrain relative sea level (RSL) evolution during the MIS 5e highstand and younger stages. Uranium-thorium dating of speleothem deposition phases places the maximum highstand RSL at 8.90 ± 0.6m a.s.l., as indicated by the near-horizontal upper limit of Lithophaga boreholes measured for along a ∼3.5 km coastal cliff section. Geochronological data show that RSL fell more than 6 m before ∼120 ka, suggesting a duration of the Last Interglacial highstand significantly shorter than proposed in some previous studies. Modelling shows that the RSL trend predicted by the ICE-5G and ICE-6G ice-sheet simulations is consistent with our data, but requires an additional significant reduction of both Greenland and Antarctic ice sheets to match the height of the local maximum highstand if no correction for tectonics is applied. Reconciling field data and models requires an earlier and likely shorter duration of the MIS 5e highstand. This suggests that our new data can constrain global ice-volume variations during the penultimate deglaciation, as well as glacial inception at the end of the Last Interglacial. According to our chronology, there is no local evidence of higher-than-present-day sea levels after 120 ka.