Three-dimensional (3D) reconstruction is a powerful tool to investigate complex neuroanatomical organizations. 3D models are often generated by piling up registered segmentations carried out on serial sections labeled by histological means. However, these models suffer limitations (incompleteness and lack of statistical representativity), which can be overcome by model averaging and fusion. These operations require an appropriate reconstruction environment allowing the simultaneous processing of several data sets. This paper describes the first release of Free-D, a software designed for the reconstruction of 3D models generated from stacks of serial sections, in the perspective of model averaging and fusion. A unique graphical user interface integrates the 3D reconstruction tools. Several large stacks (tens of gigabytes) including hundreds of images having heterogeneous characteristics (size, resolution, depth, etc.) can be simultaneously processed, thus complying to most encountered experimental situations. This first version of Free-D constitutes the required environment for the future integration of the averaging and fusion algorithms currently developed in our group and illustrated here with preliminary results.