Global change models predict not only a steady increase in temperatures, but also an increase in the occurrence of hot and cold extremes. Organisms' responses to thermal extremes will depend on species-specific traits and the degree of within-species variation (among populations), with populations from warmer latitudes often predicted to have higher thermal tolerance than populations from colder latitudes. The evolution of population-specific responses, however, can be limited by gene flow that homogenises populations. Unfortunately, most studies only examine a single thermal extreme (hot or cold) and do not compare multiple populations from multiple species in a common experimental design, thus limiting our ability to predict the winners and losers of climate change. Here, we fill these knowledge gaps with a study of the survival of Littorina littorea, L. obtusata, and L. saxatilis – marine snails with varying dispersal potential – collected at two latitudes. Snails were lab-acclimated for several weeks before undergoing exposures to extreme heat, extreme cold, or ambient conditions, and individual mortality was recorded after each exposure. In line with common predictions, we observed that the degree of population divergence in survival under thermal extremes was negatively related to dispersal potential, and that populations from the colder latitude generally had higher survival to sub-freezing temperatures. Contrary to common predictions, however, we observed greater survival after extreme heat in populations from colder latitudes than in their warmer-latitude counterparts, a pattern known as countergradient variation. This experiment highlights counterintuitive responses to thermal extremes, emphasising that colder-latitude populations could experience population growth under more extreme climates due to higher survival at both hot and sub-freezing thermal extremes.