The purpose of this study was to develop a Mathematics test instrument testlet model for a classroom assessment at elementary school. Testlet Model is a group of multiple choice question acquiring similar information with different grade of responses model. This research was conducted in East Lombok, Indonesia. The design used was research development model. The study involved 17 doctoral students as peer reviewers, 19 teachers as testlet preparation workshop participants and 711 students for testing the testlet mathematical test instrument. Data collection techniques used were documentation techniques, interviews and questionnaires. Expert validation results were analyzed using Content Validity Index (CVI). Analysis of the trial results was using items response theory with a graded response model of two logistic parameters. Independent sample t test was used for comparison of tests. The research developed a Mathematics test instrument testlet model successfully for a classroom assessment consisting of the testlet model test design, grading guide, test instrument preparation guide and mathematics test instrument testlet model or an accurate and reliable classroom assessment. The result of the analysis shows that there was different result between assessments using testlet compared to multiple choices scoring, with t value equal to 7.864.