How populations respond to novel or rapidly changing environmental conditions depends, at least in part, on how fitness effects of alleles change with the environment. Variation in the effects of alleles across environments could result in greater genetic variation, and potentially facilitate rapid adaptation enabling evolution to keep up with the current environmental changes. However, we know relatively little about how allelic effects vary, and in particular how such variation affects evolvability across environmental gradients. To address this question, we introduced new alleles, unsorted by historical selection, into a population of zebrafish, Danio rerio, using the mutagen ENU. We characterised the effect of this new mutational variance on swimming performance across a thermal gradient to test the hypothesis that mutational effects vary with temperature in such a way as to increase genetic variation in environments that are increasingly different to the historically experienced environment. Swimming speed was genetically correlated across temperatures, with the most mutational variance associated with faster/slower swimming speeds across all temperatures. This suggests selection could act on this axis of variation to remove slow individuals from the population. We did not observe support for the hypothesis that mutational effects were smallest in the ancestral temperature, but rather that mutational effects of swimming speed vary little over an ecologically relevant temperature range in zebrafish.