Students' morality and character are greatly influenced by Islamic religious education. As a result, the learning process must be carried out as effectively as possible at home and in the classroom using a variety of methods, strategies, and models. However, in practice, a lot of PAI classes continue to use traditional approaches that center on the instructor. Low student active participation results from this. Innovation in the use of learning models that can boost student engagement and comprehension is required to get past this. The purpose of this study is to explain how the differentiated learning-based think talk write (TTW) learning model is implemented and examine how the model affects student learning results. With a Quasi Experiment Nonequivalent Group Design research design, a quantitative methodology is employed. Tests, documentation, and observation were used as data collection methods. The findings demonstrated that the differentiation-based TTW model significantly and favorably improved learning outcomes. The fact that the experimental class's average posttest score increased to 87.14, higher than the control class's (61.86), is proof of this. Furthermore, the independent sample t test computation results indicate a Sig. (2-tailed) of 0.000, indicating that Ha is accepted and Ho is rejected. Thus, this model is declared effective in improving students' PAI learning outcomes.