The purpose of this study is to determine whether the think pair share learning model with a  problem based learning approach is effective based on students' self-efficacy. This research is a quantitative research in the form of quasi-experimental research using a pretest-posttest control group design. The sample of the study was the students of X MIPA class of MAN Salatiga in the academic year 2018/2019, the sample was determined using the Cluster Random Sampling technique. Data was collected using a self-efficacy questionnaire consisting of 20 statements. The results of the study show that: (1) the results of the independent sample t test mean students' self-efficacy obtained a value of 0.035, (2) the average N-Gain score of self-efficacy of students who learned by using the think pair share model with a problem based learning approach of 0.36 while the average N-Gain score of self-efficacy of students who get conventional learning is only 0.23. Thus it can be concluded that the think pair share learning model with a problem based learning approach is effective based on the self-efficacy of X MIPA class of MAN Salatiga students in the academic year 2018/2019.