There are a range of plant modelling approaches including process-based models, L-systems or functionalstructural models, simplified big leaf models, models aimed at exploring the impacts of genes, through to simple model plants that do little more than provide an uptake term in a water balance. Each model is designed to address different problem domains using a diverse range of software and algorithms.At a glance these models appear to be very different. However, there is similarity between many of the components within these models. For example, the basic abstraction of a plant canopy into individual phytomers can be very similar in models that differ greatly in complexity. Within many plant models there is also a general concept of resource supply and demand, and internal plant competition for these resources. At the same time, there can be complementarity in the strengths and weaknesses of these different modelling approaches. The strengths of one modelling approach are often the weakness of other approaches. Functional-structural plant models have strengths in modelling the light environment and low-level plant processes, but are not necessarily designed for modelling larger scale farming systems problems. Farming systems models have a long history in modelling the water, carbon, nutrient and management aspects of cropping systems but are struggling to deal with important questions arising from manipulation of plant architecture through genetics or planting geometry.The requirements for plant models are increasing rapidly to deal with the impacts of changing climates on crops and weeds, environmental impacts on plant growth and resource capture, emergent properties of plant genetic traits or to assist in designing production systems to address issues of food security whilst minimizing environmental impact. With such acceleration in model requirements we must learn from the plant breeders who have dramatically accelerated plant improvement and look for ways to apply these principles to improving our virtual plants.In this paper we compare the processes of traditional and modern plant improvement and relate them via analogy to the development of virtual plants. Ongoing cross-fertilisation of ideas, evolution and evaluation is required and will require two main things. Software methods are needed to allow the exchange of code or algorithms. Just as importantly, increased dialog between researchers is required to allow the lessons from one model domain or paradigm to be transferred to another.