The Eocene-Oligocene Transition (EOT) marks the shift from greenhouse to icehouse conditions at 34 Ma, when a permanent ice sheet developed on Antarctica. Climate modeling studies have recently assessed the drivers of the transition globally. Here we revisit those experiments for a detailed study of the southern high latitudes in comparison to the growing number of mean annual sea surface temperature (SST) and mean air temperature (MAT) proxy reconstructions, allowing us to assess proxy-model temperature agreement and refine estimates for the magnitude of the pCO2 forcing of the EOT. We compile and update published proxy temperature records on and around Antarctica for the late Eocene (38-34 Ma) and early Oligocene (34-30 Ma). Compiled SST proxies cool by up to 3°C and MAT by up to 4°C between the timeslices. Proxy data were compared to previous climate model simulations representing pre- and post-EOT, typically forced with a halving of pCO2. We scaled the model outputs to identify the magnitude of pCO2 change needed to drive a commensurate change in temperature to best fit the temperature proxies. The multi-model ensemble needs a 30 or 33% decrease in pCO2, to best fit MAT or SST proxies respectively, a difference of just 3%. These proxy-model intercomparisons identify pCO2 as the primary forcing of EOT cooling, with a magnitude (-200 or -243 ppmv) approaching that of the pCO2 proxies (-150 ppmv). However individual model estimates span -66 to -375 ppmv, thus proxy-model uncertainties are dominated by model divergence.