Understanding how schooling structures can mitigate or amplify achievement gaps is a central issue in education from both an equity and an efficiency perspective (Hanushek & Woessmann, 2008; Van de Werfhorst & Mijs, 2010). Highly selective and tracked systems provide an analytically powerful setting for studying these processes (Van de Werfhorst & Mijs, 2010), and Hungary is a particularly informative case because international assessments place it among the most highly stratified OECD systems (OECD, 2023). Subsequent analyses have also shown that the Hungarian education system is distinctive in that it selects students not only between schools, but also between classes within schools (Csapó, Molnár, & Kinyó, 2009). Existing analyses are mostly based on cross-sectional data collections, and do not cover the eight years (ISCED 1 and 2) of primary education in Hungary. The present analyses fill this niche and analyzes how performance gaps between schools and classrooms evolve from Grades 2 to 8 (ISCED 1-2) using data from the Hungarian Educational Longitudinal Program (N=6,231; 155 schools, 287 classes; Csapó, 2014). Reading comprehension was measured via the eDia online assessment system (Csapó & Molnár, 2019) using anchored tests with increasing proportions of inferential and evaluative items at higher grade levels. A Rasch model was used to place scores on a common developmental scale (EAP/PV reliability=0.88). Missing data were handled following established methodological recommendations (e.g. Wijesuriya et al., 2025) and implemented multiple imputation with 100 datasets generated via a random-forest chained-equations algorithm. Between-school differences were statistically significant at all four measurement points and showed a steady increase over time (Grade 3: F=9.58; Grade 4: F=12.41; Grade 6: F=14.53; Grade 8: F=15.44; p<.001). Class-level disparities across schools were likewise significant and grew across the school years as well (Grade 3: F=7.36; Grade 4: F=8.72; Grade 6: F=9.97; Grade 8: F=10.43; p<.001). However, class-level differences within schools remained largely stable (Grade 3: F=4.04; Grade 4: F=3.58; Grade 6: F=3.65; Grade 8: F=3.53; p<.001), indicating persistent within-school stratification from early primary grades onward. The findings demonstrate that substantial performance stratification is already present by Grade 3 at both school and classroom levels. These achievement gaps increase between schools and stay stable between classes within schools throughout primary education. This poses a serious challenge to equity objectives in Hungary and in countries with similarly selective systems, and should draw attention to the relevance of policy interventions aimed at mitigating educational selection.