The core element in German school inspections is the evaluation of the quality of classroom teaching using questionnaire-supported expert ratings. The criteria for performance measurement are in most cases based upon research on effective teaching and are, therefore, oriented on the process-product paradigm of school effectiveness research. Complex models for describing the quality of classroom teaching are missing as well as benchmarks for the assessment of and standards for the appraisal of performance, which allow leading personnel in school and education administration to answer questions based on the results of evaluations and to facilitate evidence-based governance and teaching development. This article suggests the development of a multi-level model for classifying the performance in classroom teaching using comparative empirical research. The data (N = 2240) is derived from a sample of the Hamburg School Inspection and the model has been developed using Item Response Theory. The results show that it is possible to develop a model which allows the conversion of quantitative measurements into performance levels with a narrative description of the content which is typical at each level. Furthermore, it can be demonstrated that the model is valid for different types of schools and that single lesson sequences can be discriminated at the margins of the scale as well as in the middle. Nevertheless, the structure of data is found to be multi-dimensional, so that a uni-dimensional scaling procedure may lead to a loss of information and inaccurate estimates. It can be shown, however, that the expected bias is of little significance and that strong correlations between the sub-dimensions of the model can be found. Thus, it can be assumed that using a uni-dimensional model of performance levels to describe the quality of classroom teaching may be a reasonable and empirically tenable approximation of the multidimensional model. Finally, a multi-level model is presented, which is developed using a proficiency scaling.