Modeling of the Antarctic stratospheric polar vortex from the Coupled Model Intercomparison Project phase 3 (CMIP3) to phase 6 (CMIP6) is evaluated in this study. On average, a wide coverage of warm biases appears in the Antarctic stratosphere, which is greatest in the early CMIP and is gradually diminished in the two later CMIPs with the number of models producing QBO increasing. Four metrics of the Antarctic stratospheric polar vortex are assessed for three generations of CMIPs. Biases such as the overly weak strength, the overly large aspect ratio and the westward drifted vortex centroid are commonly shared across the CMIPs. While with improvements of the model resolution, model top, interactive chemistry and physical process, the intermodel spread narrows generation by generation, especially for high-top models than low-models in the simulation of vortex area. Further, Intermodel spread of Antarctic stratospheric vortex is obviously associated with the bias of austral winter sea surface temperature (SST). Speci cally, a warm SST bias in the southern oceans, including southern Indian Ocean and southern Niño 1 + 2 regions is signi cantly linked to the weaker vortex strength and the westward-displaced vortex centroid, which can be partly attributed to the modifying of the upward propagations of planetary waves in tropical and extratropical oceans. The strengthened relationships in the focused regions further con rms the importance of the SST simulation for the stratosphere vortex simulation. In general, despite biases of the polar vortex existing across CMIPs, marked progresses have been achieved for most models.