Abstract. During the Last Glacial Maximum (LGM, ∼ 23 000 to 19 000 years ago), the Patagonian Ice Sheet (PIS) covered the central chain of the Andes between ∼ 38 to 55° S. Existing paleoclimate evidence – mostly derived from glacial landforms – suggests that maximum ice sheet expansions in the Southern Hemisphere and Northern Hemisphere were not synchronized. However, large uncertainties still exist in the timing of the onset of regional deglaciation and its major drivers. Here we present an ensemble of numerical simulations of the PIS during the LGM. We assess the skill of paleoclimate model products in reproducing the range of atmospheric conditions needed to enable an ice sheet growth in concordance with geomorphological and geochronological evidence. The resulting best-fit climate product is then combined with records from southern South America offshore sediment cores and Antarctic ice cores to drive transient simulations throughout the last 70 ka using a glacial index approach. Our analysis suggests a strong dependence of the PIS geometry on near-surface air temperature forcing. Most ensemble members underestimate the ice cover in the northern part of Patagonia, while tending to expand beyond its constrained eastern boundaries. We largely attribute these discrepancies between the model-based ice geometries and geological evidence to the low resolution of paleoclimate models and their prescribed ice mask. In the southernmost sector, evidence suggests full glacial conditions during marine isotope stage 3 (MIS3, ∼ 59 400 to 27 800 years ago), followed by a warming trend towards MIS2 (∼ 27 800 to 14 700 years ago). However, in northern Patagonia, this deglacial trend is absent, indicating a relatively consistent signal throughout MIS3 and MIS2. Notably, Antarctic cores do not reflect a glacial history consistent with the geochronological observations. Therefore, investigations of the glacial history of the PIS should take into account southern midlatitude records to capture effectively its past climatic variability.