The model architecture might be ten old, but it seems it was trained recently, so the title is a bit misleading.