This study aimed to know if students who work in heterogeneous (HET) pairs have significantly better writing ability than those who experience working in homogenous (HOM) pairs. This study involved two intact classes that consist of 40 EFL students taking the Intermediate Reading and Writing course in the English Education Department in one of the large private universities in Indonesia. This study employed a causal-comparative design and lasted for twelve meetings including pretest and posttest. The two groups of HET pairs and HOM pairs experienced collaborative writing activities following the steps of the Genre-based Approach. The data were collected through writing pre-test and post-test. The data were analyzed using descriptive statistics and Mann Whitney to compare the students' post-test scores. The findings show that both high and low proficiency students who experienced collaborative writing in homogenous proficiency pairings have better writing ability than those who experienced collaborative writing in heterogeneous proficiency pairings. This indicates that pair collaboration can support language learning more optimally when there are no large proficiency gaps among pairs.