The aim of the research was to determine the effect of learning models on mathematics achievement viewed from the student creativity. The learning model compared were RME with problem solving, RME with problem posing on conventional. The type of the research was a quasi-experimental research. The population was the students of Junior High School in Surakarta city on academic year 2012/2013. The samples of this research were taken by using the stratified cluster random sampling technique. The samples consisted of 269 students who were divided into 90 students in the first experiment class, 90 students in the second experiment class, and 89 students in the control class. The result of research showed that: (1) RME with problem solving provided better learning achievement than RME with problem posing and the conventional model. The RME learning model with problem posing provided better learning achievement than conventional one, (2) the students having high creativity had better learning achievement than those having medium and low creativity. The students having medium creativity had better learning achievement than did those having low creativity, (3) students taught by using RME with problem solving having high creativity level had better learning achievement than those having medium and low creativity level and students having medium creativity level had learning achievement as good as the low creativity level. Students taught by using RME with problem posing and conventional model had equal learning achievement in each level of creativity, and (4) students having high creativity level taught by using RME with problem solving had better learning achievement than those taught by using RME with problem posing and conventional model. Students having high creativity taught by using RME with problem posing and conventional model had equally good learning achievement. Students having medium and low creativity level had equal learning achievement in each learning model.