||Product modelling has received a lot of attention in the last decennium and is now growing into a successful means to support design and production processes, also in the area of building and construction. Collaboration through data exchange and model integration are coming within reach for all participants in the building process. However, the applicability of the current approaches in product modelling for architectural design is still very limited. It is the nature of architectural design to give much importance to issues such as uniqueness and diversity in relation with architectural style. Particularly in the earlier stages of the design process, not just technical but also cultural issues play an important role. Standardisation and predefined methodologies of design are not generally appreciated during early design, when ambiguity and a dynamic way of handling design information is often considered very important. The success of computer support for architectural design therefore depends on how well it supports ambiguity and a dynamic handling of design information. This criterion for successful design support systems seems to oppose the need for standardisation and classification that is felt so strongly in the later stages of the building process.The paper describes and discusses three long-term, independent research projects that are being carried out in three European universities: the BAS·CAAD project , the IDEA+ project , and the VR-DIS project *. While their initiatives were independent and the developments are not formally related, these three projects show strong similarities in terms of objectives, conceptual approach, and methodology. The paper demonstrates that these parallel research projects are paving a new way for the development of design support systems, allowing architects to profit from the benefits of product modelling technologies and enabling integration of early design stages in the complex process of building design and construction.The common objectives of the projects are identified in detail. One of the major issues is schema evolution, or the necessity for a design model to be conceptually adaptable as design proceeds and more information is becoming available or design decisions are reversed. It is also recognised that no assumptions can be made about design methods, and that design information models must support, for instance, both spatial design and design that starts from building elements. Design concepts such as space and user activity play an important role in early design stages and must take a central role in design models as well.Approaches to achieve these objectives can be positioned in the force-field of two pairs of opposite characteristics of design information models. The first pair is (1a) maximum consistency and optimal data exchange through rigidly predefined typologies, versus (1b) maximum flexibility and extensibility of typologies in the conceptual schema. The second pair distinguishes approaches based on (2a) domain independent concepts from those based on (2b) specific domain concepts. The paper discusses the position in these force-fields of each of the three projects, which also clarifies their individual theoretic bases for information modelling. Although these theoretic bases are different in the three projects, common for all three is the object orientation of their approach and, more importantly, the effort to disconnect the identification of objects from the properties of objects. This appears to be an effective means to facilitate flexibility. Also common to the three projects, but elaborated very differently in each of them, is the capability of user-defined extensions to the conceptual schema. Both these issues of flexibility and extensibility are discussed in detail in the paper.Finally, the paper summarises the individual conclusions drawn in three PhD theses reporting intermediary and final results from the projects. This leads to the final discussion of the potentials of schema evolution for the integration of early design stages in the product modelling process. As a basis for the next generation of architectural design support tools, dynamic information models can be expected to deliver an important contribution to the rationalisation of architectural design and are an important next step in solving the conflict between computer tools and designers’ creativity.