AbstractUsing 16 CMIP5/6 models with a spontaneously generated quasi-biennial oscillation (QBO)-like phenomenon, this study investigates the impact of the QBO on the northern winter stratosphere. Eight of the models simulate a QBO with a period similar to that observed (25–31 months), with other models simulating a QBO period of 20–40 months. Regardless of biases in QBO periodicity, the Holton–Tan relationship can be well simulated in CMIP5/6 models with more planetary wave convergence in the polar stratosphere in easterly QBO winters. This wave polar convergence occurs not only due to the Holton–Tan mechanism, but also in the midlatitude upper stratosphere where an Elissen–Palm (E-P) flux divergence dipole (with poleward E-P flux) is simulated in most models. The wave response in the upper stratosphere appears related to changes in the background circulation through a directly excited meridional–vertical circulation cell above the maximum tropical QBO easterly center. The midlatitude upwelling in this anticlockwise cell is split into two branches, and the north branch descends in the Arctic region and warms the stratospheric polar vortex. Most models underestimate the Arctic stratospheric warming in early winter during easterly QBO. Further analysis suggests that this bias is not due to an overly weak response to a given QBO phase, as the models simulate a realistic response if one focuses on similar QBO phases. Rather, the model bias is due to the too-low frequency of strong QBO winds in the lower stratosphere in early winter simulated by the models.