Carcass composition and classification is a key determinant of beef cattle farmers’ incomes and is mainly assessed by subjective visual notation (CHTAX system in Switzerland), which is prone to inter-operator variability. The present study aimed to evaluate the use of dual-energy X-ray absorptiometry (DXA) scans, a promising non-subjective technique, to predict the tissue composition of beef carcass from half-carcass or 11th rib cut scan. Eighteen crossbreds bulls (♀ Brune Suisse x ♂ Angus (n=6), Limousin (n=6) or Simmental (n=6)) were slaughtered at either 460-470 kg (n=6) or 520-530 kg (n=12) of body weight. Half-carcass and 11th rib cut were scanned by DXA and half carcass was further anatomically dissected. Simple linear regressions were performed aiming at estimate half-carcass muscle and adipose tissue proportions obtained by dissection from DXA lean and fat proportions measurements, respectively on half-carcass or on 11th rib cut. Muscle proportion is predicted using DXA half-carcass lean proportion with an accuracy of R2 = 0.64 and residual coefficient of variation (rCV) = 3.0%. Adipose tissue proportion was predicted with an accuracy from DXA half-carcass fat proportion of R2 = 0.81 and rCV = 9.7%. When prediction of tissue composition was made by DXA scan of 11th rib cut, precision of equation decreased slightly (R2 = 0.62 and rCV = 3.1% for muscle and R2 = 0.69 and rCV = 12.5% for adipose tissue). Adding the crossbreed effect on the intercept of multiple regressions improve the accuracy of estimates of carcass composition from DXA scan.