Resistance emerging after ceftazidime or cefepime therapy was investigated in a peritonitis model. Mice were given a peritoneal challenge (10* cfu plus talcum) and treated by either antibiotic (50 mg/kg/dose, which produced similar antibiotic concentrations in peritoneal fluid in both cases). After one or three doses, resistance never developed in Serratia marcescens or Citrobacter freiouiii infections. After Enterobacter cloacae and Pseudomonas aeruginosa challenge, ceftazidime selected more resistance (21/36 cases) than did cefepime (1/36 cases). In mice challenged with resistant strains selected by ceftazidime therapy, cefepime (six doses) successfully treated 7/18 E. cloacae infections but 0/18 P. aeruginosa infections; ceftazidime was never effective. Neither cefepime nor ceftazidime cured mice infected with the resistant strain selected by cefepime. MICs were poor predictors of further emergence of resistance in mice inoculated with strains classified as susceptible, but antibiotic-containing agar gradients plated with a high inoculum (10* cfu) allowed better prediction. In selected clinical situations, cefepime may be preferable because it may be associated with less frequent emergence of resistance.