Accurate in vivo estimation of body composition of beef cattle is of great interest in terms of recording nutritional status or determination of optimal time for slaughter. The aim was to estimate the body and carcass anatomical compositions of crossbreed beef bulls (n=36, 477 to 530 kg body weight (BW), mothers Brown Swiss and sires Angus, Limousin and Simmental, the most widespread in Switzerland) using ultrasound (US) measurements. Subcutaneous adipose tissue (AT) and muscle (MU) thickness were measured by US (9L-RS Linear-Sonde (4.0 MHz); GE Healthcare) at 4 specific body locations (ischium-sacrum, lumbar, dorsal, flank) 3 to 7 days before slaughter. After slaughter, empty body weight was determined after removal of digestive content, the weight of visceral AT (omental, mesenteric, perirenal and pericardiac) was recorded, as well as the weights of carcass AT and MU after anatomical dissection of one half-carcass. Relationships between BW and US measurements as predictive variates, and empty body and carcass AT and MU proportions were analysed by linear regressions using GLM and PHREG procedures (SAS 9.4). Empty body (456.8±20.5 kg) contained 13.9±2.7 (9.5 to 21.9)% AT, hot carcass (288.6±16.0 kg) contained 14.1±3.0 (9.0 to 23.6)% AT, and 64.3±3.0 (58.2 to 70.8)% MU. The best single estimate was AT dorsal thickness, including crossbreed type as a fixed effect, for empty body AT% (R2=0.72, residual standard deviation (rSD)=1.5%, P<0.05), carcass AT% (R2=0.70, rSD=1.7%, P<0.05) and carcass MU% (R2=0.58, rSD=2.0%, P<0.10). BW combined with crossbreed type did not explained empty body AT% nor carcass AT% or MU% (P>0.20). Neither dorsal MU thickness, nor US measurements at the ischium-sacrum, lumbar or flank (R2≤0.58) locations were good estimators of empty body or carcass composition. In conclusion, the dorsal AT thickness measurement by US gave an accurate estimator of empty body and carcass compositions and seems therefore suited for the in vivo follow-up of their dynamics.