Biomass production, soil water extraction, and water-use efficiency (WUE, kg dry matter (DM)/ha.mm growing-season water use) of tropical, summer-growing and temperate, winter-growing forage legumes suited to short-term rotations with crops were compared over several growing seasons in southern Queensland. Tropical legumes lablab (Lablab purpureus cvv. Highworth and Endurance), burgundy bean (Macroptillium bracteatum cvv. Cardarga/Juanita mix), and butterfly pea (Clitoria ternatea cv. Milgara) were compared with forage sorghum (Sorghum spp. cv. Silk and cv. Sugargraze). Temperate legumes snail medic (Medicago scutellata cv. Sava), lucerne (Medicago sativa cv. UQL-1), sulla (Hedysarum coronarium cv. Wilpena), and purple vetch (Vicia benghalensis cv. Popany) were compared with forage oats (Avena sativa cv. Taipan/Genie). Production and WUE of winter legumes was highly variable, with oats producing more biomass than the legumes, except in 2009 where oat establishment was poor. In years with good establishment, WUE of oats (14–28 kg DM/ha.mm), snail medic (13–25 kg DM/ha.mm), and sulla (12–20 kg DM/ha.mm) were similar, but the production and WUE of vetch were generally lower (6–14 kg DM/ha.mm). Sulla dried the soil profile by 60–100 mm more than the annual species, but less than lucerne. Summer legumes, burgundy bean, and lablab performed similarly, although always produced less biomass and had lower WUE than forage sorghum. Lucerne extracted more water and maintained a drier profile by 70–150 mm and had lower WUE (<10 kg DM/ha.mm) than burgundy bean or lablab (9–30 kg DM/ha.mm). Of the legumes tested, burgundy bean and lablab seem the most likely to be profitably integrated into subtropical cropping systems. Further evidence of the rotational benefits provided by these legumes is required before they will be favoured over the perceived reliability and higher productivity of annual grass-type forages.