The purpose of the present study was to shorten the 122-item STEM Interest Complexity Measure which is a valid measure of interests towards increasingly complex tasks covering the domains of numeric, symbolic, spatial, and STEM-ideas within the Science, Technology, Engineering, and Mathematics (STEM) areas. In a sample of 418 undergraduates enrolled in STEM majors, polytomous IRT using the Graded Response Model was applied. Items that best discriminated across interest trait levels were retained. The 43-item shortened version of the measure offered equivalent validity estimates with vocational criteria of STEM-GPA, STEM-major satisfaction, and intentions to pursue increasingly complex STEM occupations. The study contributes to IRT applications in scale refinement and to producing a practical tool in vocational interest assessments.