Many Augmented and Mixed Reality applications are based on two libraries: OpenGL is used for rendering and ARToolKit is used for marker recognition. The ARToolkit library is great for rapid prototyping of AR/MR applications. The library is very easy to use and it hides the complexity of marker recognition. In the AMIRE (Authoring Mixed Reality) project, a European founded AR project, we follow up the aim of ARToolKit consistently: the AMIRE approach is to offer well-established gems (software solutions) and components for a faster prototyping of AR/MR applications. Each content user should be able to develop his/her own AR/MR application without any computer graphics skills. Therefore, one of the primary goals of AMIRE is to find well-established solutions of current AR/MR applications. One of the solution is the ARToolKit library, which is used in numberless AR/MR applications. But which library should be used for rendering? Can we use a high-level graphics AN together with ARToolKit? Which graphic library would be the best for farther AR/MR applications? Should developers use Direct3D/OpenGL or should we propose a high-level graphics API, like Open Inventor, OpenGL Performer, OpenSG, or Open SceneGraph? High-level graphics APIs have a number of advantages as opposed to low-level graphics APIs. They include: A set of loaders (e.g. model and texture loaders); A scenegraph concept; Modern object oriented design; High performance (optimized rendering, view frustum culling, small object culling, Level of Detail nodes, etc.). One problem still remains: How difficult is the usage of ARToolKit, originally based on OpenGL, in combination with a high-level graphics API like OpenSG? For the AMIRE project we tested two different APIs: Open SO and Open SceneGraph.