Even though IDEs are often a central tool when learning to program in CS1, many teachers still lean on paper-based exams. In this study, we examine the “test mode effect” in CS1 exams using the Rainfall problem. The test mode was two-phased. Half of the participants started working on the problem with pen and paper, while the other half had access to an IDE. After submitting their solution, all students could rework their solution on an IDE. The experiment was repeated twice during subsequent course instances. The results were mixed. From the marking perspective, there was no statistically significant difference resulting from the mode. However, the students starting with the paper-based part tended to make more errors in their code, but after the computer-based reworking phase, they matched or exceeded the level of the students who started with the computer-based phase. We also discuss the reliability of automatic assessment that is based on a unit test suite that was develo ped for the purposes of this study.