The extent to which the size of a modular stellarator reactor may be reduced is investigated by means of an analytic model of the reactor. The various means employed include varying the blanket/shield thickness, the power output and the wall loading. An optimum design is found, the major radius of which tends to be insensitive to changes in these quantities, although a decrease in the power output leads to a rather smaller decrease in reactor dimensions, as would be expected. Varying the plasma beta at fixed ι2 or, alternatively, increasing the rotational transform per field period, may, however, allow configurations with fewer field periods to be accessed which have a substantially smaller major radius than the 'standard case' adopted. The magnetics of various configurations required by the model are checked by field line following and the performance claimed is shown to be relatively easily achieved. Costs of two promising configurations are estimated; these designs are found to be highly competitive with other conceptual fusion reactors in terms of cost, as well as requiring less advanced technologies.