<![CDATA[Deep learning represents a transformative agenda in twenty-first-century education, emphasizing cognitive engagement, conceptual connections, metacognitive reflection, knowledge transfer, and collaboration. However, evidence from secondary-school classroom practices, particularly across subjects, remains limited. This study examined the implementation of deep learning in Grades 10 and 11 at SMAN 19 Bone and developed a model grounded in findings. An embedded mixed-method design was employed, with a dominant qualitative approach supported by observations. Participants included eight teachers of Mathematics, Indonesian Language, Science, and Social Studies and 64 students. Data were collected through observations, semi-structured interviews, document analysis, and a five-indicator deep learning rubric. Quantitative analyses included descriptive statistics, one-way ANOVA, post hoc tests, and an independent-samples t-test. Results indicated high to very high levels of deep learning. Indonesian Language achieved the highest score, while Social Studies recorded the lowest. Cognitive engagement was the strongest indicator, whereas metacognitive reflection was the weakest. Significant differences were found across subjects and between Grades 10 and 11. Qualitative findings showed that problem-based, inquiry-based, and project-based learning enhanced cognitive engagement, although reflection was not systematically integrated. The study proposed the Deep Learning Cross-Disciplinary Framework, positioning reflection as a moderating variable with implications for curriculum, teacher development, assessment, and instruction.]]>
