The architecture of living cells is difficult to describe and communicate; therefore, realistic computer models may help their understanding. 3D models should correspond both to qualitative and quantitative experimental data and therefore should include specific authoring tools such as appropriate visualization and stereological measures. For this purpose we have developed a problem solving environment for stereology-based modeling (PSE-SBM), which is an automated system for quantitative modeling of cell architecture. The PSE-SBM meets the requirement to produce models that correspond in stereological and morphologic terms to real cells and their organelles. Instead of using standard interactive graphing tools, our approach relies on functional modeling. We have built a system of implicit functions and set operations, organized in a hierarchical tree structure, which describes individual cell organelles and their 3D relations. Natural variability of size, shape, and position of organelles is achieved by random variation of the specific parameters within given limits. The resulting model is materialized by evaluation of these functions and is adjusted for a given set of specific parameters defined by the user. These principles are explained in detail, and modeling of segments of a muscle cell is used as an example to demonstrate the potential of the PSE-SBM for communication of architectural concepts and testing of structural hypotheses.