This study investigated the ability of two models to represent glucose kinetics in the basal steady state and during an intravenous glucose tolerance test (IVGTT). Six young nonobese male subjects were studied after an overnight fast. Two bolus injections of [U-13C]glucose were given 150 min apart, the first without and the second together with concomitant injection of unlabeled glucose. [3-3H]glucose was constantly infused throughout the study and served to provide an independent means for evaluation of system responses. A linear time-invariant three-compartmental model and the two-compartment time-variant model proposed by Caumo and Cobelli were used to interpret measured time courses of [U-13C]glucose and to reconstruct endogenous glucose production and glucose removal. The ability of the two models to describe the glucose tracer time course was comparable. Simulation studies showed that the two-compartmental time-variant system better predicted measured [3-3H]glucose concentration profiles than did the three-compartmental time-invariant model. However, endogenous glucose production and the integral of excess glucose removal over basal during the IVGTT derived from the two models were almost identical.