A B S T R A C T Quantification of the enhancement in cleavage fracture toughness of ferritic steels following warm pre-stressing has received great interest in light of its significance in the integrity assessment of such structures as pressure vessels. A Beremin type probability distribution model, i.e., a local stress-based approach to cleavage fracture, has been developed and used for estimating cleavage fracture following prior loading (or warm pre-stressing, WPS) in two ferritic steels with different geometry configurations. Firstly, the Weibull parameters required to match the experimental scatter in lower shelf toughness of the candidate steels are identified. These parameters are then used in two-and three-dimensional finite element simulations of prior loading on the upper shelf followed by unloading and cooling to lower shelf temperatures (WPS) to determine the probability of failure. Using both isotropic hardening and kinematic hardening material models, the effect of hardening response on the predictions obtained from the suggested approach has been examined. The predictions are consistent with experimental scatter in toughness following WPS and provide a means of determining the importance of the crack tip residual stresses. We demonstrate that for our steels the crack tip residual stress is the pivotal feature in improving the fracture toughness following WPS. Predictions are compared with the available experimental data. The paper finally discusses the results in the context of the non-uniqueness of the Weibull parameters and investigates the sensitivity of predictions to the Weibull exponent, m, and the relevance of m to the stress triaxiality factor as suggested in the literature. a = crack length (mm) W = ligament (mm) a/W = crack/ligament ratio B = specimen thickness (mm) B 0 = reference thickness (mm) B/B 0 = thickness correction ratio i = order number (ascending) of a specific specimen tested in a group (i = 1, . . . , N ) β = 'shape parameter' exponent for toughness-based distribution m = Weibull exponent N = total number of specimens tested under the same conditions (sample size) K 0f = reference fracture toughness (MPa √ m) K f = fracture toughness after WPS (MPa √ m) K min f = minimum (threshold) fracture toughness (MPa √ m)