Code generation is widely used to make software development more efficient and less prone to human errors. A significant use case of code generation is processing of Domain-Specific Languages (DSLs) and Domain-Specific Models (DSMs). Sometimes, it is desired to generate semantically equivalent or similar functionality to different languages to better support multiple platforms and achieve better reuse in the tooling. For example, it is convenient if a single tool supports code generating from a DSM to either Java or C#. There has been relevant research on using modeling and model transformations for code generation to multiple platforms. The Model-Driven Architecture (MDA) inherently supports multi-platform code generation based on models. Nevertheless, the MDA standard is a high-level general framework that includes standards, notions and principles but does not specify more concrete methods or workflows about their efficient adoption. Our research focuses on the efficient and practically usable application of MDA principles to generate multi-platform code. This paper reports on our results on multi-platform code generation and the difficulties that we are about to addressed in future research. The approach and the challenges presented in the paper are useful for tool developers, such as developers of DSLs, who generates code for several platforms.