Abstract-The problem in this research is the lack of student's Mathematical Problem Solving (MPS) ability. To overcome this problem, researchers tested the learning strategies that hone skills that are called meta-cognitive. A Meta-cognitive learning strategy is a habituation learning to control the thought process through a series of questions meta-cognition. One strategy meta-cognition is through think-aloud that can be done in a cooperative setting either in pairs (pair) as well as groups of four (square). The purpose of this study is to compare and analyze the MPS between students who received learning Metacognition Think-Aloud in cooperative (Think-Pair-Share) (MTA-P), Meta-cognition Think-Aloud in a cooperative (Think-PairSquare) (MTA-S), and Direct Instruction (DI). This study is the mixed methods with quasi-experimental design posttest only the students of class VIII Madrasah Tsanawiyah (MTs)/Boarding School (PP) in Pekanbaru. The data was collected with MPS test, MPS task analysis and interviews. Data was analyzed by descriptive statistics and comparative test. Based on the analysis, it can be concluded that: 1) There is a difference of MPS between students overall based strategy (MTA-P, MTA-S, DI), based on the School system (MTs, PP), school rankings (A, B), based on their General Math Ability/KMU (high, medium, low), and gender male (M), Female (F). MPS of students at MTA-P is better than MPS of students on other two strategies; 2) There is interaction between learning strategy based on schools ranks in MPS students; 3) The most dominant indicator in the process of thinking of students in MPS is to understand the problem, and the least is to re-check the obtained solution.