Scientific models are not, and are not supposed to be, true. Rather they represent their targets by fitting their targets. Fitting is a matter of degree. Some philosophers, such as Giere and the early Nersessian, maintain that fitting depends on similarity. A model fits its target to the extent that it is similar to the target in relevant respects. I argue that exemplification rather than mere similarity is the relation that connects a model with its target. I explicate exemplification, showing that it is a symbolic relation between an exemplar and some of its features. I then show how a model (1) exemplifies features it shares with its target, (2) highlights the significance of those features, (3) affords epistemic access to them in the target, and (4) thereby affords an understanding of the target system.