The ability to solve mathematical problems is a very important part that must be owned and developed by students in learning mathematics. However, generally Indonesian students' abilities are very low in understanding complex information and problem solving. Therefore, this article aims to know: 1) the increase in Mathematical Problem Solving Ability (MPSA) of students who get Problem Based Learning (PBL) better than students who get conventional learning with a scientific approach is reviewed overall and based on KAM levels; 2) find out the interaction between PBL and conventional models with KAM level to MPSA; 3) the difference in MPSA increase of students who get PBL learning in terms of KAM level. The research method used was a quasi-experimental research design with a pretest-posttest control group design. The sample of 6 grade students of SMAN 2 Bangkinang Kota class XI totaling 183 students. The instrument in this article is a MPSA test using t-test, two-way ANOVA test and one-way ANOVA test. The results in this study were: 1) the increase in MPSA of students who received PBL learning was better than students who received conventional learning in terms of overall and KAM; 2) there is a significant interaction between PBL and conventional models with KAM level to MPSA; and 3) there is a difference in the increase in MPSA of students who get PBL learning in terms of KAM level significantly.