Abstract: A Monte Carlo Potts model was developed to simulate the recrystallization process of a cold-rolled ultra-thin grain-oriented silicon steel. The orientation and image quality data from electron backscatter diffraction measurements were used as input information for simulation. Three types of nucleation mechanisms, namely, random nucleation, high-stored-energy site nucleation (HSEN), and high-angle boundary nucleation (HABN), were considered for simulation. In particular, the nucleation and growth behaviors of Goss-oriented (011<100>) grains were investigated. Results showed that Goss grains had a nucleation advantage in HSEN and HABN. The amount of Goss grains was the highest according to HABN, and it matched the experimental measurement. However, Goss grains lacked a size advantage across all mechanisms during the recrystallization process.