The objective of this study is to investigate cognitive performance in a first-episode psychosis sample, when stratifying the interaction by cannabis use and familial or non-familial psychosis. Hierarchical-regression models were used to analyse this association in a sample of 268 first-episode psychosis patients and 237 controls. We found that cannabis use was associated with worse working memory, regardless of family history. However, cannabis use was clearly associated with worse cognitive performance in patients with no family history of psychosis, in cognitive domains including verbal memory, executive function and global cognitive index, whereas cannabis users with a family history of psychosis performed better in these domains. The main finding of the study is that there is an interaction between cannabis use and a family history of psychosis in the areas of verbal memory, executive function and global cognition: that is, cannabis use is associated with a better performance in patients with a family history of psychosis and a worse performance in those with no family history of psychosis. In order to confirm this hypothesis, future research should explore the actual expression of the endocannabinoid system in patients with and without a family history of psychosis.