This research aims to determine differences in learning outcomes IPA-Physics students taught using models problem Solving with learning Model TPS mode and conventional learning. Hypothesis (H1) formulated in this study is the "Third Problem solving learning model, Think-Pair-Share and learning Konvensioanal learning outcomes IPA-physics are not the same". This research is quasy experiment and design of the study is a post test only control group design. The study population was all students of class VII MTsN Tarusan, South Coastal District first semester of academic year 2016/2017. Sample were taken using cluster random sampling technique. Retrieved class VII B as the experimental class I (Problem Solving), a class VII D as the experimental class II (Think-Pair-Share) and class VII C as the control class (conventional learning).The average results of studying physics at the cognitive aspects of the experimental class I (Problem Solving) is 80.46 in the experimental class II (Think-Pair-Share) was 78,27,14 and the control class method was 70 (Conventional). Based on hypothesis testing, f count > f table (6,49 > 3.10) so that H0 and H1 accepted, meaning all three learning model provides science-physics learning outcomes are not the same. After the test of the hypothesis H0 decision, to determine which model of learning most good, do multiple comparison test (multiple comparison) with Scheffe test. the results of the data analysis Scheffe test showed Fs> F table (11,68 > 9.55). So we can conclude there are differences in all three classes on the material characteristics of the sample substance and physical and chemical changes.