For suitable production planning in an open pit mine, and also to carry out the optimisation process during the mining production cycle, prediction of fragment size distribution is required prior to carrying out the blasting operation. For this purpose, many prediction models, such as empirical models, were presented. The proficiency of these models differs in various conditions and mines. Therefore, in this paper the most important models, including SveDeFo, Kuz-Ram, CZM, Modified Kuz-Ram in 2005, Kuznestsov-Cunningham-Ouchterlony (KCO) and also Gheibie et al, were examined by the information of 20 blasts in Gol-Gohar iron ore mine. Fourteen images were taken on average for each blast. For image processing analysis Split-Desktop software was used. After examination, it was understood that none of the available empirical fragmentation models had presented acceptable performance, except the Kuz-Ram model for magnetite ore, which had somewhat acceptable results. However, this model for blasting patterns in hematite presented unsuitable results. Examination and comparison of two important distribution functions, Rosin-Rammler and Swebrec, were carried out and it was observed that the performance of Swebrec for fines was far better than Rosin-Rammler. This result for the upper part of fragment size distribution curve, coarse, is the reverse.CITATION:Jahani, M and Taji, M, 2015. Comparison of empirical fragmentation models at the Gol-Gohar iron ore mine, in Proceedings 11th International Symposium on Rock Fragmentation by Blasting, pp 707-714 (The Australasian Institute of Mining and Metallurgy: Melbourne).