With regard to the moderate performance of German students in international large-scale assessments, one branch of German science education research is concerned with the construction and evaluation of competence models. Based on the theory-driven definition of competence levels, these models imply a correlation between the complexity of a question or a problem, its difficulty and the cognitive demands that are required to answer or solve it. The aim of the study was to apply two competence models in order to analyse the complexity of questions that chemistry teachers use to promote learning in class as well as to compare the results of this analysis. Two model-based coding schemes were constructed and evaluated on the basis of interrater reliability before analysing the teacher questions in 40 chemistry lessons. The results show that between 60 and 65% of the questions refer to low complexity levels. Although there is a considerable correspondence between the results (69.2%), neither model can be considered as redundant. These findings are discussed with regard to the level and the development of students' skills in science.