The current research concentrated on some areas of improvement in the HEMFAST code with the goal of quantitative furniture fire predictions. The bench scale data processing routines is revised to take advantage of the recent upgrades in the cone calorimeter apparatus and of the consequential new scaling relationships for the materials burning rates and combustion products. The prediction accuracy of the furniture fire modules is improved by restructuring the code for better time integration of virgin surface temperature flame front positions, and local burn history. Other improvements include a model of convective heat transfer on burning areas and a new soot formation model. A theoretical relationship between the specific soot extinction area exiting the flame and the flaming soot absorption coefficient was developed and partially validated. The cone calorimeter database of several furniture materials was processed into a form that can be used by the furniture fire modules for arbitrary sized fires.