The output from a design process of high precision and reliable haptic devices for surgical training like bones and teeth is a complex design. The complexity is largely due to the multi-criteria and conflicting character of the functional requirements. These requirements include high stiffness, large workspace, high manipulability, small inertia, low friction, and high transparency. The requirements are a basis for generating design concepts. The concept evaluation relies to a large extent on a systematic usage of kinematic, dynamic, stiffness, and friction models. The design process can benefit from a model-based and simulation driven approach, where one starts from an abstract top-level model that is extended via stepwise refinements and design space exploration into a complete realization of the system. Such an approach is presented and evaluated through a test case where a haptic device, based on a Stewart platform, has been designed and realized. It can be concluded, based on simulation and experimental results that the performance of this optimally designed haptic device satisfies the stated user requirements. This indicates that the methodology can support the development of an optimal haptic device. However, more test cases are needed to further verify the presented methodology.