The consistency of the changes in extreme warm spell probabilities among the different observational datasets and model simulations examined suggests that they are robust regional aspects of global warming associated with atmospheric circulation changes. This highlights the need for climate models to represent not just the mean regional temperature signals but also the changes in subseasonal temperature variability associated with global warming. However, current climate models (both CMIP3 and CMIP5) generally underestimate the magnitude of the changes in the atmospheric circulation and associated temperature variability. A likely major cause of this is their continuing underestimation of the magnitude of the spatial variation of tropical SST trends. By generating an overly spatially bland tropical SST warming in response to changes in radiative forcing, the models spuriously mute tropically-forced teleconnections around the globe, and hence mute changes in temperature variability. This leads to overemphasized impacts of the mean temperature changes, and underemphasized impacts of the changes in subseasonal temperature variability, on the altered risks of extreme warm spells in those climate model simulations.