<![CDATA[Mathematical problem-solving ability is one of the essential competencies that students must possess in mathematics learning. However, the mathematical problem-solving ability of eleventh-grade students at SMA Negeri 1 X Koto has not yet reached an optimal level. This study aimed to analyze and compare whether the mathematical problem-solving ability of students who learned using the Deep Learning approach was better than that of students who learned through the conventional approach. This study was a quasi-experimental study employing a Posttest-Only Control Group Design. The population consisted of eleventh-grade students of SMA Negeri 1 X Koto in the 2026/2027 academic year. Samples were selected using purposive sampling technique, with class XI.F1 (33 students) as the experimental class and class XI.F2 (35 students) as the control class. Data were collected through a final test of mathematical problem-solving ability and analyzed using normality, homogeneity, and t-tests. The results showed that the average score of the experimental class (55.76) was higher than that of the control class (36.38). Furthermore, the t-test results (P-value = 0.000 < 0.05) confirmed that students who learned using the Deep Learning approach had significantly better mathematical problem-solving ability than those who learned through the conventional approach. Therefore, the Deep Learning approach had a positive effect on students' mathematical problem-solving ability. ]]>
