ABSTRACT This study discusses how to improve students' abilities after using a problem-based learning model in mathematics learning material. This type of research is quantitative research. Quantitative research consists of experimental methods and survey methods. In this study using the experimental method. The experimental method used to find interaction. The research design used was a quasi-experimental design with class 2 research taken from class VIII SMP in one of the cities of Sukabumi. The research sample used cluster random sampling technique. The research design used in this study was the design of the pretest-posttest control group. The instrument used in this study is essay questions as many as 6 items which have been declared valid and reliable. Based on the prerequisite test of the data produced by the research data, it is not normally distributed so that further testing is carried out, namely the mann-withney test. From the data analysis, the average posttest value of the experimental students with the problem based learning learning model was 18,24 and the average score of the posttest control class students without using the problem based learning learning model was 15,17. There is a significant difference in student learning outcomes, this is evidenced by the mann-withney test so this study rejects H0 and accepts H1. Thus, it can be concluded that there is an increase in learning outcomes in mathematical punishment abilities of students who use learning by using problem-based learning models on the material to get a flat side space.