This research aims to find out whether there are differences and improvements in the mathematical problem-solving skills of students who apply the teams games tournament (TGT) learning model with students who apply the conventional learning model. This research uses experimental quantitative methods. The experimental design used in this study is Quasi Experimental Design. The population in this study are students of class VIII SMP IT Bina Insan Mulia. The class level that was sampled for this study was taken by purposive sampling. The instrument used to collect data is in the form of test questions. Data analysis techniques using the t test that was previously carried out by the data analysis requirements test are the normality test obtained by normally distributed data and the homogeneity test obtained by homogeneous data. From the results of the research, it can be concluded that the average pretest score of students’ mathematical problem solving ability is 50.4 and after being given a TGT learning model, the average posttest student's mathematical problem solving ability is 97.1. As for the control class, it can be concluded that the average pretest score of students’ mathematical problem solving ability is 25.6 and after being given a conventional learning model, the average posttest of students' mathematical problem solving ability Is only 57.4. Thus, the TGT learning model has differences and improvements to the mathematical problem-solving skills of students in grade VIII of SMP IT Bina Insan Mulia