Abstract: This research aimed to find out: (1) which one is better in giving the mathematical word problem solving ability among learning models of jigsaw with problem posing approach, learning models of jigsaw, or direct instructional models, (2) which one is better in giving the mathematical word problem solving ability among studentsâ level of confidence, students having high, medium or low level, (3) in each level of confidence, which one is better in giving the mathematical word problem solving ability among learning models of jigsaw with problem posing approach, learning models of jigsaw, or direct instructional models and (4) in each learning models, which one is better in giving the mathematical word problem solving ability among studentsâ level of confidence, students having high, medium or low level. This research was a quasi-experimental research with 3Ã3 factorial design. The population of the research was all grade VII students of Junior High School (SMP) in Metro, Lampung in academic year 2013/2014. The samples were chosen by using stratified cluster random sampling. The instruments that were used to collect the data were the test and questionnaire of studentâs level of confidence. The technique of analyzing the data was two-ways ANOVA with unbalanced cells. The results of research showed as follows. (1) Learning models of jigsaw with problem posing approach had mathematical word problem solving ability as good as learning models of jigsaw, learning models of jigsaw with problem posing approach had better mathematical word problem solving ability than direct instructional models, and learning models of jigsaw had mathematical word problem solving ability as good as direct instructional models. (2) The students having high level of confidence had mathematical word problem solving ability as good as those having medium level of confidence. The students having high level of confidence had better mathematical word problem solving ability than those having low level of confidence. The students having medium level of confidence had mathematical word problem solving ability as good as those having low level of confidence. (3) In each level of confidence, learning models of jigsaw with problem posing approach had better mathematical word problem solving ability than direct instructional models. (4) In each learning models, the students having high level of confidence had better mathematical word problem solving ability than those having low level of confidence.Keywords : Jigsaw with Problem Posing Approach, Direct Instructional Models, Â Â Â Mathematical Word Problem Solving Ability, Confidence.