This study examined individual differences in spatial abilities of architecture students. Students at different educational levels were assessed on spatial ability tests that varied in their domain-specificity to architecture, with the hypothesis that larger differences between beginner and advanced students will emerge on more domain-specific tests. We also investigated gender differences in test performance and controlled for general reasoning ability across analyses. In a cross sectional study, master students (N = 91) outperformed beginners (N = 502) on two novel tests involving perspective taking and object composition, as well as on a standardized visualization of cross-sections test, but not on a standardized mental rotations test. Longitudinally (N = 117), spatial performance improved after the first bachelor year on visualization of cross-sections, object composition and mental rotation. Although both genders showed higher spatial test performance with increased experience, male students outperformed females across tests and levels of education. The results overall confirmed improvements in spatial performance during architecture studies, with partial support for the domain-specificity hypothesis. A gender gap among advanced students calls for further examining architecture-specific spatial thinking.