Different proxies for sea surface temperature (SST) often exhibit divergent trends for deglacial warming in tropical regions, hampering our understanding of the phase relationship between tropical SSTs and continental ice volume at glacial terminations. To reconcile divergent SST trends, we report reconstructions of two commonly used paleothermometers (the foraminifera G. ruber Mg/Ca and the alkenone unsaturation index) from a marine sediment core collected in the southwestern tropical Indian Ocean encompassing the last 37,000 years. Our results show that SSTs derived from the alkenone unsaturation index (UK′37) are consistently warmer than those derived from Mg/Ca by ~2–3°C except for the Heinrich Event 1. In addition, the initial timing for the deglacial warming of alkenone SST started at ~15.6 ka, which lags behind that of Mg/Ca temperatures by 2.5 kyr. We argue that the discrepancy between the two SST proxies reflects seasonal differences between summer and winter rather than postdepositional processes or sedimentary biases. The UK′37 SST record clearly mimics the deglacial SST trend recorded in the North Atlantic region for the earlier part of the termination, indicating that the early deglacial warming trend attributed to local summer temperatures was likely mediated by changes in the Atlantic Meridional Overturning Circulation at the onset of the deglaciation. In contrast, the glacial to interglacial SST pattern recorded by G. ruber Mg/Ca probably reflects cold season SSTs. This indicates that the cold season SSTs was likely mediated by climate changes in the southern hemisphere, as it closely tracks the Antarctic timing of deglaciation. Therefore, our study reveals that the tropical southwestern Indian Ocean seasonal SST was closely linked to climate changes occurring in both hemispheres. The austral summer and winter recorded by each proxy is further supported with seasonal SST trends modeled by Atmosphere–ocean General Circulation Models for our core site. Our interpretation that the alkenone and Mg/Ca SSTs are seasonally biased may also explain similar proxy mismatches observed in other tropical regions at the onset of the last termination.