Abstract: Model based testing is a crucial dimension in software testing. However, with the increase of model scale, model based test case generation is becoming more and more arduous. Extended Finite State Machine (EFSM) has been widely used in industry, which is extended from Finite State Machine (FSM), and can depict the dynamic behavior of software system more accurately. EFSM based test case generation mainly includes two parts: test transition paths generation and test data generation that covers the test transition paths, in which search based technology is adapted in test data generation. In order to improve the efficiency of test case generation in large-scale EFSM models, EFSM slicing based test case generation and test case compensating are proposed based on the previous research on EFSM dependence analysis and slicing for non-termination of EFSM models. Two case studies are introduced to show that model slicing based test case generation is more accurate in feasible path generation and test intensity improvement. In this paper, the experiments on 7 standard EFSMs are conducted, and the results show that all of the test case generated from slice can be used in the original model, and in most cases, test case generation efficiency on slice is higher than that on the original model.