Multi-unit auctions are being used by public agencies to allocate resources and to purchase services. These auctions resolve the lumpy bid problem inherent in single-bid auctions by allowing bidders to submit a supply or demand schedule. However, the choice of pricing formats for multi-unit auctions is controversial. Neither economic theory nor laboratory experiments depict a complete picture of how alternative pricing formats perform in terms of budgetary or allocative efficiency. This paper constructs an agent-based computational model to compare uniform, discriminatory and generalised Vickrey pricing formats under different degrees of competition and heterogeneity in the bidder population.