Worldwide, teachers are expected to engage their students in authentic practices, like scientific modeling. Research suggests that teachers experience challenges when integrating modeling in their classroom instruction, with one explanation that teachers themselves lack the necessary modeling competence. Currently, theoretical conceptualizations structure the modeling competence into three dimensions: meta-modeling knowledge, modeling practice, and modeling products. While each of these dimensions is well researched on its own and the three dimensions are commonly expected to be highly positively related, studies investigating their specific relationships are widely lacking. Aiming to fill this gap, the present study investigated the meta-modeling knowledge, modeling practice, and modeling products of 35 secondary preservice biology teachers engaging in a black box modeling task. Data were collected with an established pen-and-paper questionnaire consisting of five constructed response items assessing meta-modeling knowledge and by videotaping the participants engaging in the black box modeling task. Herein, the three dimensions of modeling competence were operationalized as five variables including decontextualized and contextualized meta-modeling knowledge, complexity, and homogeneity of the modeling processes and a modeling product score. In contrast to our expectations and common assumptions in the literature, significant relationships between the five variables were widely lacking. Only the complexity of the modeling processes correlated significantly with the quality of the modeling products. To investigate this relationship further, a qualitative in-depth analysis of two cases is presented. Implications for biology teacher education are discussed.