The learning model is a pattern or steps in learning that are applied so that the expected learning outcomes are achieved more effectively and efficiently. The model chosen must be by the characteristics of the class. Besides that, students have a tendency not to listen to the teacher's explanation if they are less involved in the process. Then, the carousel Feedback learning model can present the missing element, which is more involved students in the learning process. This study aims to determine the effect of the Carousel Feedback learning model on students' mathematical problem-solving skills. The research method used is quantitative with a pseudo-experimental design (nonequivalent control group design). The research was conducted on VIII grade students at MTs Nuraniyah, involving 40 students in the experimental class and 42 students in the control class. The material used was Two-Variable Linear Equation System (SPLDV). The results showed that the mathematical problem-solving ability of students taught with the Carousel Feedback model was higher than the conventional model. The average post-test score of the experimental class was 76,58, which was higher than the control class score of 68,14. N-gain analysis showed an increase of 0,45 (medium category) for the experimental class and 0,24 (low category) for the control class. The t-test resulted in a significant value of 0,00 > 0,001, indicating a significant effect on the learning outcomes.