Radiofrequency (RF) ablation is increasingly used to treat cancer because it is minimally invasive. However, it is difficult for operators to control precisely the formation of coagulation zones because of the inadequacies of imaging modalities. To overcome this limitation, we previously proposed a model-based robotic ablation system that can create the required size and shape of coagulation zone based on the dimensions of the tumor. At the heart of such a robotic system is a precise temperature distribution simulator for RF ablation. In this article, we evaluated the simulation accuracy of two numerical simulation liver models, one using a constant thermal conductivity value and the other using temperature-dependent thermal conductivity values, compared with temperatures obtained using in vitro experiments. The liver model that reflected the temperature dependence of thermal conductivity did not result in a large increase of simulation accuracy compared with the temperature-independent model in the temperature range achieved during clinical RF ablation.