This research is based on the ability of mathematical connections of students who are still relatively low, seen from the results of PISA 2012 and 2015. The low yield is because the learning has not been well facilitated, as well as the habituation of teachers to develop indicators of mathematical connection capability. Viewing the learning of mathematics that has the concept of learning systematic and has a relationship with each concept demands for every human being can think logically, systematically, and rationally. So that the selection of learning models that can develop the ability, one of which multiliteration model. This learning consists of four learning cycles: setup, explore, share and discuss, and presenting. The research was conducted on the students of Grade V Elementary School 2013 with the theme of Theme 8 on building simple spaces (cubes and beams). The purpose of this research is to know the difference of improvement and comparison of mathematical connection ability using multiliteration model with ordinary learning (scientific). This research is quasi-experimental with nonequivalent control group design. The research was conducted at SDN 036 Ujungberung and SDN 050 Cibiru. Based on the results of mathematical connection test data processing, the difference of students' mathematical connection ability is based on the average score of post control class score of 68.52 and the experimental class is 78.61 after receiving treatment with the difference of 10,09. The improved quality of mathematical connection ability of the experimental class and control class is in the medium category. Based on gain analysis data of mathematical connection ability of student of experiment class is better than control class with difference 0,185. So that the multiliteration model becomes one of the alternative learning models that can be used in an effort to improve students' mathematical connection ability.