Supraglacial lakes (SGLs) are widely distributed across Antarctica and play an important role in modulating surface energy balance through albedo feedback, promoting ice-shelf disintegration via hydrofracture, and influencing ice dynamics. Existing SGL monitoring studies mainly rely on narrow-swath satellite data, such as Landsat and Sentinel-2, resulting in discontinuous observations with relatively long revisit intervals and limiting the ability to capture the rapid evolution of supraglacial hydrological processes. Here, an 8-day Antarctic SGL fraction dataset spanning 2000–2023 is presented. The dataset is generated by integrating high-temporal-resolution Moderate Resolution Imaging Spectroradiometer (MODIS) imagery with high-spatial-resolution Sentinel-2 data within a machine-learning framework. The dataset reveals that Antarctic SGLs are highly dynamic and short-lived, with approximately 83 % of lakes exhibiting mean persistence rates below 5 %. The multi-year mean maximum SGL area is estimated at 4,103 ± 1,479 km<sup>2</sup>. Clear spatial heterogeneity in peak timing is further revealed, with SGL extent peaking approximately one week earlier in West Antarctica than in East Antarctica and about two weeks earlier than on the Antarctic Peninsula. Spatially, approximately 65 % of SGLs are located within 10 km of grounding lines and are closely associated with blue-ice and exposed rock areas. These long-term, high-temporal-resolution observations provide a valuable basis for investigating the spatiotemporal variability of Antarctic SGLs and their associated impacts. The dataset is publicly available at <a href="https://doi.org/10.5281/zenodo.19936100" target="_blank" rel="noopener">https://doi.org/10.5281/zenodo.19936100</a>.