AbstractThe importance of models and modeling in science education is well‐recognized, yet there exists significant polysemy among these terms within the literature. This ambiguity often leads to confusion, particularly regarding whether modeling represents an expected student performance, an instructional strategy to promote such performance, or both. Moreover, the construction of models has been depicted as both the objective of modeling and a distinct phase within modeling‐based instruction. Additionally, the expression of models has often been overlooked despite its significance as a crucial modeling practice. In an endeavor to shed light into these complexities associated with modeling in science education, this paper pursues a twofold aim. First, it theoretically presents and justifies the Instruction Performance Modeling (IPM) cycle, drawing on numerous previous contributions to the field, as a practical and specific instructional tool designed to clarify some problematic concepts both regarding modeling instruction and modeling practice. Second, it provides empirical evidence regarding the type of modeling performance exhibited by students involved in instruction guided by the IPM. This study applies discourse analysis to the multimodal productions of preservice teachers attending a lab‐based workshop on the topic of flotation. The main findings reveal that students' modeling performance, while exhibiting certain patterns such as the Introductory pattern or the Evaluation‐Revision one, predominantly manifests as a disorganized sequence of modeling practices. This result is consistent with certain precedents in the modeling literature but contrasts with the expected outcomes of well‐established approaches like Generation‐Evaluation‐Modification. Furthermore, the study aims to highlight the rich, meaningful, and productive modeling practices occurring in instructional scenarios guided by the IPM cycle.