To determine how current predictors of carcass and cut yield perform in the Australian lamb industry a large study was undertaken using diverse genotypes. Carcass measurements and yield data were obtained for 591 lambs representing two sexes (cryptorchids, ewes). The lambs were sired by a selection of Texel (T), Poll Dorset (PD), Border Leicester (BL) and Merino (M) rams crossed with Border Leicester×Merino (BLM) and Merino (M) ewes giving six genotypes (T×BLM, PD×BLM, T×M, PD×M, BL×M, M×M). The percentage yield of saleable meat, hindquarter, loin and a range of trimmed cuts were predicted using models based on hot carcass weight, EUROP conformation score, m. longissimus thoracis et lumborum area (LL area) in combination with three different measures of carcass fatness: (1) the depth of muscle and fat tissue from the surface of the carcass to the lateral surface of 12th rib, 110 mm from the midline (GR); (2) fat depth over the LL at the deepest part of the muscle (FATC); and (3) fat depth at the 5th rib, 110 mm from the midline (FAT5). FAT5 was identified as the most accurate and robust (stable) fat measurement for the prediction of saleable meat yield in a production system based on diverse genotypes. Using GR as a fatness measurement led to overestimation of fatness level and underestimation of yield in T sired lambs. Conformation score after carcass weight and any of the fat measurements improved (P<0.01) the prediction of the yield of saleable meat, hindquarter, loin and all trimmed lamb cuts with the exception of the ribloin (rack). LL area added significantly (P<0.01) to the prediction of the yield of saleable meat, hindquarter and trimmed lamb cuts, but not for the midloin and neck fillet at a constant carcass weight and fatness. Large proportions of the variation in all yields were unaccounted for, even after the inclusion of either conformation or LL area. Predictions based on carcass weight and any of the three different fat measurements were not very accurate. Although the addition of LL area as a predictor significantly (P<0.01) increased the accuracy of the predictions the results suggest that the current description system for lamb carcases in Australia needs to be re-examined.