Pastures sown to lucerne (Medicago sativa L.) with a perennial non-legume could increase feed supply relative to traditional pastures based on subterranean clover (Trifolium subterraneum L.). Such mixtures might also be preferable to pure lucerne pastures, which are prone to weed invasion. Yield and water-use efficiency (harvested dry matter per unit evapotranspiration) of mixtures with lucerne or subterranean clover were compared a field experiment established under rainfed conditions at Hamilton, Victoria. Soil moisture and dry matter production were measured over 2 years. Treatments included chicory (Cichorium intybus L.), cocksfoot (Dactylis glomerata L.) and tall fescue (Festuca arundinacea Shreb.) with either lucerne or subterranean clover; pure lucerne; and phalaris (Phalaris aquatica L.) and perennial ryegrass (Lolium perenne L.) with only subterranean clover. In the second year, dry matter production from lucerne mixtures exceeded that of equivalent mixtures with subterranean clover in spring, summer and winter. In spring, the lucerne component continued producing for longer than the clover component through its use of deeper stored soil water, and in summer, lucerne continued to grow slowly after the grass component had entered a drought-induced dormancy. In winter, the contribution from the lucerne component complemented, rather than competed with, that from the non-legume component. Water-use efficiencies during winter–spring ranged from 4 kg ha^–1 mm^–1 for chicory–clover to 27 kg ha^–1 mm^–1 for a fescue–lucerne mixture, and during summer–autumn from nil for cocksfoot–clover to 13 kg ha^–1 mm^–1 for a fescue–lucerne mixture. This study demonstrates that lucerne-based mixtures can increase forage supply per unit water use relative to traditional pastures based on subterranean clover.