Free-form façades with bent glass have become increasingly popular in recent times. As bent glass is stiffer against out-of-plane loads, it can result in thinner glass and lower embodied carbon. A promising new technique is to cold bend thin glass plates with a stiff structural edge into a hyperbolic paraboloid (hypar), and to subsequently lock the corners to create a self-contained, self-stressed system. In this research, the physical bending process of insulated glazing units (IGUs) with a particular local instability phenomenon is investigated. This instability is hypothesised to be delayed by stiffening the edges of the plate, which is done here by using 30x30 mm GFRP profiles as spacers. These were bonded to the glass using Dow 993 silicone adhesive. Four IGUs of 1.5x1.5 m were produced, three with 4 mm fully toughened glass, and one with 1.1 mm chemically toughened glass. In a series of experiments, the panels were supported on two opposite corners, and pulled down on the others. A numerical model was developed to predict the outcome of the experiments, as well as the behaviour of the panel under wind load, though the latter was not verified with experiments. In this finite element model, the glass plates were modelled as 2D flat shell elements. The spacers were also modelled with these elements, by using them as the walls of the square hollow profile. The silicone joints were modelled as 3D solids, using the Mooney-Rivlin material model. These parameters were tested by successfully modelling a similar panel that was tested at the University of Cambridge. In both the model and the experiments, it was found that with the sizes used it was difficult or impossible to get close to a hypar. Due to the small thickness of the glass, one of the diagonals would always be straight or mostly straight throughout the loading process, making the panel not resemble a hypar. The bottom plate of the 4 mm panels broke at a corner displacement of around 150 mm and a total load of 2.6 kN, and the top plate around 200 mm and 0.7 kN. The bottom plate of the 1.1 mm also broke first, at a corner displacement of 120mm and a total load of 1.4 kN. The top plate broke at a much higher corner displacement of almost 400mm, when the loaded diagonal was fully straight. This resulted in the top plate breaking into long, thin rods of over a meter long. It was also found that the top and bottom plates would make contact around 50 mm corner displacement in the 4 mm panels, and around 30 mm corner displacement in the 1.1 mm panel. The numerical model could predict this contact and the overall behaviour of the panel until a corner displacement of 60 mm. From the experiments and the model it was concluded that the glass was too thin for the size of the panels and the applied edge stiffening. Changing the parameters (e.g. thin glass in smaller sizes or higher edge stiffness) could result in a viable product.