Test methodologies for large embedded systems fail to reflect the test process as a whole. Instead, the test process is divided into independent test levels featuring differences like the functional abstraction levels, but also similarities such as many functional test cases. Desirable instruments such as test front-loading feature a considerable test effort and test cost reduction potential, but their efficiency suffers nowadays from the strict separation of the test levels and the consequent lack of appropriate mechanisms for reusing tests across test levels. Multi-level test cases have shown to provide the means for a seamless test level integration based on test case reuse across test levels. This paper extends this concept by introducing multi-level test models which are capable of systematically integrating different functional abstraction levels. From these models, we can derive multi-level test cases that are executable at different test levels. With this novel approach, multi-level testing benefits from the principles of model-based testing while the requirements for providing multi-level capabilities to any test models are analyzed and described.