Abstract With the advent of Software-Intensive and Cyber-Physical Systems, heterogeneous formalisms are used to model different parts of the system. Co-simulation is often used to simulate the heterogenous models simultaneously with exchange of information. One emerging industry standard in this field is the Functional Mock-up Interface. This standard defines the interface for such a co-simulation, using so called Functional Mock-up Units (FMUs). These FMUs are zip-files containing XML-files and C-code. The C-code consists of the model and its solver. Nowadays solvers are coded instead of being explicitly modeled. However, this does not allow straightforward analysis or detection of e.g. algebraic loops and optimization possibilities. Explicitly modeling the solvers might offer a solution to overcome these limitations, since this allows for the use of model-driven engineering techniques, like model transformations. The paper will present a method to generate FMUs from Causal Block Diagram models, more specific Simulink® models, with explicit modeled ODE solvers. The execution performance is compared between FMUs with explicitly modeled solvers and FMUs with coded solvers. We conclude that modeling the solver has a significant positive impact on the run-time efficiency of the generated FMU.