Compared to other psychological constructs, such as intelligence, creativity usually is a less valid predictor for scholastic achievement. The majority of the studies from which these results are derived apply Ordinary Least Squares (OLS) methods that assume random sampling of respondents and, hence, do not take into account the possibility of between-classroom variance that is due to clustered data structures. In this study, data from the standardization sample of the Berlin Structure of Intelligence Test for Youth: Assessment of Talent and Giftedness (BIS-HB) were analyzed using multilevel modelling techniques. To test the hypothesis that the impact of creativity on GPA may vary between different classrooms, a multilevel model incorporating explanatory variables on two levels (level 1: students and level 2: classrooms) was specified. The role of reasoning ability was also investigated. The results allow for a more detailed interpretation of the role of different variables in the context of predicting scholastic achievement. More specifically, it could be shown that especially the predictive power of creativity changes across classrooms, indicating that some teachers value creativity in their students more than others