Subtropical marine low cloud sensitivity to an idealized climate change is compared in six large‐eddy simulation (LES) models as part of CGILS. July cloud cover is simulated at three locations over the subtropical northeast Pacific Ocean, which are typified by cold sea surface temperatures (SSTs) under well‐mixed stratocumulus, cool SSTs under decoupled stratocumulus, and shallow cumulus clouds overlying warmer SSTs. The idealized climate change includes a uniform 2 K SST increase with corresponding moist‐adiabatic warming aloft and subsidence changes, but no change in free‐tropospheric relative humidity, surface wind speed, or CO2. For each case, realistic advective forcings and boundary conditions are generated for the control and perturbed states which each LES runs for 10 days into a quasi‐steady state. For the control climate, the LESs correctly produce the expected cloud type at all three locations. With the perturbed forcings, all models simulate boundary‐layer deepening due to reduced subsidence in the warmer climate, with less deepening at the warm‐SST location due to regulation by precipitation. The models do not show a consistent response of liquid water path and albedo in the perturbed climate, though the majority predict cloud thickening (negative cloud feedback) at the cold‐SST location and slight cloud thinning (positive cloud feedback) at the cool‐SST and warm‐SST locations. In perturbed climate simulations at the cold‐SST location without the subsidence decrease, cloud albedo consistently decreases across the models. Thus, boundary‐layer cloud feedback on climate change involves compensating thermodynamic and dynamic effects of warming and may interact with patterns of subsidence change.