This article expands upon the argument of a previous work which defended a variational account of scientific fictions. Specifically, I show that this understanding of scientific fictions can provide guidance for realist interpretations of scientific theories and models. Depending on a model’s variational properties, different ontological commitments are appropriate, providing a principled way for a realist to moderate her views according to the structural properties of a given model. This reasoning is then applied the Lee-Yang theory and Kubo-Martin-Schwinger statistics, two foundational models in quantum statistical mechanics. The Lee-Yang theory is analyzed in a way that permits a robust realist interpretation, whereas KMS statistics is shown to involve a use of fictions that shields the theory from confirmation and makes it inappropriate for strongly realist interpretation, without contradicting broadly realist commitments.