Cefazolin and cephradine were compared quantitatively in vitro and in experimentally infected mice. In vitro estimation of efficacy was based on growth curves of a strain of Escherichia coli O54 in trypticase soy broth. For the studies in vivo, an acute inflammation was induced in mice by injection of 5 × 106E. coli into the thigh. Antibiotics were administered after 1 hr, and concentrations in blood were followed for 2 hr; the antibacterial activity was estimated from bacterial counts made in homogenized, individual thighs. In vitro, cefazolin was 6.12 times more active than cephradine against E. coli. In vivo, for the log dose and for the log area under the curve (AUC) of concentration of free drug in serum, cefazolin was 3.15 times more active than cephradine, but for the AUC of the log concentration of free drug in serum, cefazolin was as much as 9.25 times more active than cephradine. These findings suggest that the prediction of antibacterial activity in vivo solely on the basis of in vitro data and pharmacokinetic parameters is not reliable. When the log AUC of free serum concentration is taken as representative for the potency of an antibiotic in vivo, cefazolin is less active relative to cephradine than would be expected from the activity in vitro, but when the AUC of the log free serum concentration is taken into consideration for the same comparison, cefazolin is even more active in vivo than in vitro.