<![CDATA[This study aims to determine the effect of the Project Based Learning (PjBL) model based on the deep learning approach on students' mathematical problem-solving abilities on spatial geometry material based on word problems. The problems found indicate that students still have difficulty understanding the context of the problem and determining the solution steps because previous learning tends to focus on memorizing formulas. This study uses a quantitative approach with a pre-experimental method, One-Group Pretest–Posttest Design. The research sample consisted of 20 sixth-grade students. The data collection technique used a test of 10 questions arranged based on problem-solving indicators according to Polya. Data were analyzed using a paired t-test and N-Gain test. The results showed an increase in students' average score from 56.50 in the pretest to 75.00 in the posttest. The paired t-test obtained a Sig. (2-tailed) value of 0.000 <0.05, so there is a significant effect of the application of PjBL based on deep learning on students' mathematical problem-solving abilities. The N-Gain value is 0.629 which is included in the moderate category. Improvements are seen at every stage of problem-solving according to Polya. Thus, the deep learning-based PjBL model effectively improves students' mathematical problem-solving abilities and is suitable for implementation in elementary school mathematics learning to support 21st-century critical thinking skills.]]>
