This study investigates the relationship between scholastic achievement (teacher-assigned grades) and cognitive abilities while explicitly modeling the multilevel structure of a large data set from an educational context. A multivariate analysis of composite scores in three subject domains (math and natural sciences, languages, and social sciences) was conducted using as predictors reasoning capacity, creativity, memory, and speed as well as student gender and type of school attended. The results showed meaningful differences in the relationships between the different abilities and school grades, and also significant contributions from student gender and school type. Specifically, reasoning capacity had the largest impact on achievement in math and natural sciences, whereas creativity explained more variability in performance for social sciences. Furthermore, girls outperformed boys. There was significant random variation in the intercepts at the between-classroom level, indicating that students sharing a common educational and ecological background tend to have more similar teacher-assigned grades when compared to students in other classrooms. Yet, there was no significant slope variation between the classrooms, which points to stable relationships between cognitive abilities and scholastic achievement.