An increasingly popular view holds that scientific modeling involves something akin to the imaginative construction of a fictional story along with its cast of fictional characters, not just the positing of entities (models) that yield a false but useful representation of their targets. The present paper focuses on the following problem for this view of models. If a model is a fiction how can it possibly be said to represent some aspect of the real world? How can the unreal represent the real, and in a way that allows modelers to make predictions about the real, and even explain some of its features? Call this the problem of the gap. The paper begins by motivating the fiction view of models, describing and contrasting the two most popular types of view (both based on Walton’s pretense theory of fiction), together with the way they deal with the problem of the gap and some other, related problems. I then sketch a modified version of the fiction view, one that takes on board aspects of both of these approaches by utilizing an important but under-appreciated feature of fiction, and I argue that the view provides natural solutions to this suite of problems.