The use of tests with unknown reliability and validity to measure student achievement was still widely practiced. This study examined the quality tests of conceptual understanding for college students in the basic concept of physics subject. This research was an R&D that used the 4D model. The test consisted of 2 types, namely multiple-choice and true-false tests, which were equipped with reasons. Development stages: 1) defined test grids; 2) designed tests; 3) developed tests; 4) expert validated of the content and construct; and 5) tested the validity, reliability, and level of difficulty. Analysis used the Rasch model assisted by Winstep software. The test results show that 18 multiple-choice and 17 true-false items fit the Rasch model in a range outfit MNSQ of 0.7-1.33 and ZSTD of -0,7-1,9. The reliability test is based on consistency in the range of 0.8-0.91 with good and excellent categories. The difficulty level shows three categories easy, difficult, and very difficult. And based on the results, 15 items are selected for each type of test. This selection is made by considering the similarity of competency outcomes measured by each test item. Both tests can be used for broader data collection to determine the best tests that represent students' conceptual understanding.