This study aims to (1) find out whether the proportion of students with mathematical connection abilities ≥ minimum criteria of mastery learning in CONINCON learning model with ethnomathematics nuances was more than the learning completeness limit, namely 75%; (2) find out whether students' mathematical connection abilities through CONINCON learning model with ethnomathematics nuances were better than students' connection abilities with Problem Based Learning model; (3) find out whether students' self efficacy influences mathematical connection abilities; (4) describe mathematical connection abilities through CONINCON learning model with ethnomathematics nuances viewed from students' self efficacy. The method used was a mixed method with a sequential explanatory design. In this research, 6 research subjects were taken based on the category of self efficacy in experimental group. Research results show that (1) proportion of students with mathematical connection abilities ≥ minimum criteria of mastery learning in CONINCON learning model with ethnomathematics nuances was more than the learning completeness limit, namely 75%; (2) students' mathematical connection abilities in CONINCON learning model with ethnomathematics nuances were better than students' connection abilities in Problem Based Learning model; (3) students' self efficacy has a significant influence on mathematical connection abilities of 68.7%; (4) description of students' mathematical connection abilities with categories : (a) high category of self efficacy are able to fullfill all indicators of mathematical connection abilities; (b) medium category of self efficacy can fulfill two indicators and less able to fullfill two indicators; (c) low category of self efficacy only fulfill an indicator.