In recent years, the architectural style for building Web Services called "Representational State Transfer" (REST) gained a lot of popularity in industry and academia. Since designing complex, distributed hypermedia systems still meeting all the REST constraints is a difficult task, an academic, model-driven approach based on a multi-layered metamodel was developed in order to enforce REST compliance. Apart from that, multiple REST API description languages emerged in industry, providing means to formally define the structure of an API for human (e.g. API documentation) and machine (e.g. automated creation of client/server stubs) consumption. This work aims to compare the academic metamodel with API description languages widely used in industry. As a comparison methodology, bidirectional model transformations were designed and implemented between the academic metamodel and each of the two leading API description languages, Swagger and RAML. The model transformations were evaluated with a quantitative method by applying them on real world API descriptions as well as manually evaluating the quality of the transformed models. The model transformations show that indeed various mappings can be established between model elements of different metamodels. However, there are also crucial differences which are also examined in this thesis.