This article addresses the issue of performance-appropriate marks for pupils in Austrian elementary schools at the end of fourth grade, based upon a multi-level analysis including the effect of the performance composition of classes on individual marks. The analyses were performed using data from the 2013 Learning Standards Assessment Mathematics 4 (N = 73,655) conducted by the Federal Institute for Educational Research, Innovation and Development of the Austrian School System (BIFIE). In addition to individual test performance and the performance composition of classes, gender, socio-economic status (SES), and the immigrant backgrounds of pupils were taken into account. Overall, logistic regressions reveal a relationship higher 0,60 between test performance and marks. This means that teachers were fairly successful in matching the performance rankings of primary school pupils with marks across classes and schools. However, the results also indicate the less-than-adequate evaluation of particularly low-performing and high-performing students. In addition, a positive deviation of marks compared to test performance was found for students from families with higher SES, and a negative deviation for students with immigrant backgrounds (within classes only). No systematic deviation of marks compared to test performance was found in relation to gender.