Vector graphics rendering techniques are described. Graphics processing units (GPUs) can render vector graphics images according to graphic trees having graphic leafs, each representing a graphics object (e.g., a shape) depicted in a vector graphics image. The described techniques involve generating groups of graphics objects depicted in an image such that graphics objects of a group have a same object type, e.g., shape. Transformations are determined that describe how to transform a first graphics object of a group to obtain other graphics objects of the group. The first graphics object is tessellated and a metadata buffer generated for the group having information indicative of the transformations. The metadata buffer is attached to a graphic leaf representing the first graphics object and graphic leafs representing the other graphics objects are removed from the graphic tree. The GPU renders objects by group based on the tessellated object and the metadata buffer's information.