Previous studies were able to successfully classify facets of intelligence and social-emotional competence into the SOKA (self- other knowledge asymmetry) model, using self- and peer estimates. This study examines whether differences in self- and peer estimates are due to an internal comparison of the facets of intelligence, as proposed by the GI/E model (generalized internal/external frame of reference model). Further, moderations of the internal comparison were tested. Differences in performance and the accuracy of self- and peer estimates across three types of schools were also tested, as well as the differentiation between self-concept and self-estimate. A sample of 249 pupils (on average age 17) were tested. They attended a senior high school with a focus on linguistic or mathematics or attended a senior high school for nursery education. Tests about the numerical, verbal and spatial intelligence as well as social-emotional competence were presented. Also, the pupils estimated their own performance and the performance of two other colleagues. The grades and self-concepts in Math, German, English and Geometry were also tested. Regressions were calculated to test the internal comparison of intelligence facets. MANOVAS were tested to analyze the difference in performance as well as the accuracy of self- and peer estimates across school types. The results show no internal comparison effect in estimating ones own intelligence. Differences in performances and self- and peer estimations are mostly due to gender differences. High correlations between self-concepts and self-estimates are found and raise the question, whether there can be differentiated between them. Results emphasize that objective performance tests should be preferred instead of school grades when consulting young people about their academic choices.