A coupled atmosphere–ocean model is necessary for tropical cyclone (TC) prediction to accurately characterize ocean feedback on atmospheric processes within the TC environment. Here, the ECMWF coupled global model is run at horizontal resolutions from 9 to 1.4 km in the atmosphere, as well as 25 and 8 km in the ocean, to identify how resolution impacts forecast accuracy of four observed major TCs in the Atlantic: Irma, Florence, Teddy, and Ida. Most of the resolutions used here are unprecedented for global models. GOES-16 and synthetic aperture radar (SAR) satellite images and best track data are used for atmospheric validation. Salinity and temperature observations from Air-Launched Autonomous Micro-Observer (ALAMO) floats are used to validate modeled upper-ocean response, including mixed layer deepening, sea surface cooling, and near-inertial waves in the wakes of TCs. Increasing atmospheric resolution leads to more realistic TC structure and stronger winds, significantly improving TC intensity forecasts and modestly improving track errors. Ocean resolution impacts the upper-ocean response but does not influence atmospheric forecasts for the fast-moving TCs considered here. Stronger mixing, sea surface cooling, and near-inertial oscillations are found for both higher atmosphere and ocean resolutions, provided the initial upper-ocean state is the same for the two ocean resolutions. Whether this agrees better with the ALAMO observations also depends on the realism of the initial upper-ocean state in the model, emphasizing the importance of ocean initialization for the accurate upper-ocean response. Overall, the model at all resolutions correctly predicts stronger mixing, surface cooling, and near-inertial oscillation amplitudes to the right of a TC center, as observed by ALAMO floats.