The presented study is intended to suggest the best model to predict students’ academic performance at university level. For this purpose, primary data was collected from 400 undergraduate and graduate students of eight departments of Mirpur University of Science and Technology (MUST), which were selected through stratified random sampling. CGPA is used as an indicator of students’ academic performance. Stepwise linear regression is used to select the best model to predict students’ academic performance at tertiary level. The final model selected through stepwise regression includes six variables: the student’s IQ, ownership of AC, gender, geographic location, self-study hours and ownership of fridge as significant predictors of students’ academic performance at tertiary level. IQ, ownership of assets and self-study hours are found to have a positive effect on CGPA while being male and the distance of the household to nearest market are found to have a negative effect on CGPA.