Nowadays, Monte Carlo calculations are commonly used for the evaluation of dose distributions and dose volume histograms in eye brachytherapy. However, currently available eye models have simple geometries, and main substructures of the eye are either not defined in details or not distinguished at all. In this work absorbed doses of eye substructures have been estimated for eye plaque brachytherapy using the most realistic eye model available, and compared with absorbed doses obtained with other available eye models. For this, a medium-sized tumour on the left sides of the right eye was considered. Dosimetry calculations were performed for four different eye models developed based on a literature review, and using a 12 mm Collaborative Ocular Melanoma Study plaque containing Cs,Pd, and I sources. Obtained results illustrate that the estimated doses received by different eye substructures strongly depend on the model used to represent the eye. It is shown here that using a non-realistic eye model leads to a wrong estimation of doses for some eye substructures. For example, dose differences of up to 35% were observed between the models proposed by Nogueira and co-workers and Yoriyaz and co-workers, while doses obtained by use of the models proposed by Lesperance and co-workers, and Behrens and co-workers differed up to 100 and 63% as compared to the situation when a realistic model was used, respectively. Moreover, comparing different radionuclides showed that the most uniform dose distribution in the considered tumour region was that fromCs, with a coefficient of variation of 33%. In addition, considering the realistic eye model, it was found that the radiosensitive region of the lens received more than the threshold dose of cataract induction (0.5 Gy), for all investigated radionuclides.