Modelling of structures exposed to fires is prone to be heavily influenced by uncertainties in geometrical parameters, thermal material data as well as uncertainties in the boundary conditions. Assessing the effects and influences in variations of all the uncertain parameters is often cumbersome and traditional methods are impractical thus modelling of the total uncertainty is needed. Uncertainty Quantification with deterministic sampling is one possible way ahead to accommodate and evaluating the effects of uncertainties with as few repeated simulations as possible. In this paper the uncertainties stemming from error in the input data and boundary conditions on one example of a steel beam under four point bending exposed to fire is evaluated.