AbstractThis article investigates the accuracy of various crack width prediction models and the newly proposed modified tension chord model (MTCM). A large number of experimental crack widths have been collected from the literature, including 203 specimens of reinforced concrete (RC) members subjected to bending and tension. The prediction models are described with upcoming new formulations and database validation. The modeling uncertainty is found by comparing the predicted crack widths against experimental data obtained using a log‐normal distribution. The results show that fib Model Code 2010 and MTCM provide the best crack width predictions of the collected databases; MTCM has the fewest mechanical simplifications of the investigated models and no empirical modifications for fitting towards experimental databases, in contrast to the approaches in Eurocode 2 and Model Code. However, the latter do predict the crack width to a reasonably good extent and are more suited for practical dimensioning than the MTCM. The findings in this article suggest that the MTCM should serve as a point of departure for further development of crack width calculation methods, and that it may have an extensive range of possible applications in the future.