Abstract Potential CO2 and brine leakage from geologic sequestration reservoirs must be quantified on a site-specific basis to predict the long-term effectiveness of geologic storage. The primary goals of this study are to develop and validate reduced-order models (ROMs) to estimate wellbore leakage rates of CO2 and brine from storage reservoirs to the surface or into overlying aquifers, and to understand how the leakage profile evolves as a function of wellbore properties and the state of the CO2 plume. A multiphase reservoir simulator is used to perform Monte Carlo simulations of CO2 and water flow along wellbores across a wide range of relevant parameters including wellbore permeability, wellbore depth, reservoir pressure and saturation. The leakage rates are used to produce validated response surfaces that can be sampled to estimate wellbore flow. Minima in flow rates seen in the response surface are shown to result from complex nonlinear phase behavior along the wellbore. Presence of a shallow aquifer can increase CO2 leakage compared to cases that only allow CO2 flow directly to the land surface. The response surfaces are converted into computationally efficient ROMs and the utility of the ROMs is demonstrated by incorporation into a system-level risk analysis tool.