Many strategies have been proposed for the treatment of sepsis, and most of the proposed treatment modalities have failed in clinical trials. Many of the previous treatment protocols called for blocking the activity of a single, clearly defined mediator. The underlying hypothesis was that sepsis induced a specific mediator that then caused organ injury and death. This simple, linear reasoning was frequently based on cytokines that were defined using endotoxin models of sepsis. The endotoxin models were widely used to study the pathophysiology of sepsis and were felt to adequately reproduce the full spectrum of inflammatory changes observed in patients with sepsis. Based on mortality and hematologic changes, these assumptions appeared justified. As the models were examined more closely, and directly compared with focus of infection models that more accurately portray the changes in sepsis, it became apparent that the endotoxin models did not accurately mimic the patient with sepsis. In the endotoxin models, the explosive release of cytokines into the circulating blood volume was reproducibly found regardless of the species studied (human, primate, pig, rat, or mouse). This lead to a series of anticytokine sepsis trials, all of which failed. The cytokine response in focus of infection models, such as that induced by cecal ligation and puncture, was examined and found to be more similar to that observed in patients with sepsis. When cytokine inhibitor strategies were used in the cecal ligation and puncture model, they were also generally found to lack efficacy. Compounds that have been shown to be effective at reducing mortality in endotoxin models should be re-evaluated in more clinically relevant models of sepsis.