This paper presents the German translation and replication of the Students’ Understanding of Models in Science (SUMS) instrument, aiming to assess how first-semester university students comprehend the submicroscopic level in chemistry courses. The assessment of students’ understanding is a prerequisite for improving teaching practices, particularly in addressing the persistently high drop-out rates observed in chemistry and chemistry-related programs. Employing a quantitative methodology, a sample of 181 undergraduate chemistry students was surveyed. The data were analyzed using structural equation modeling, resulting in two statistical models that demonstrated an excellent fit to the data, although no empirical preference could be established for one model over the other. Based on the investigation, framing models as exact replicas of the natural world cannot be considered an empirically meaningful dimension of understanding models in science. Additionally, the reliabilities of the latent constructs were found to be insufficiently low to establish generalizable measurements. These findings are discussed with a focus on epistemology and advocate for a stronger integration of model theory in chemistry teaching and learning. Finally, the importance of establishing a stronger connection between empirical evidence and the implementation of curricular changes in higher education is emphasized.