Grain quality determination involves important stages such as collection of the representative sample, homogenization, and dilution. The interrelation among sampling, homogenization, and working sample size is essential to the reliability of the information generated. Therefore, this work aimed to analyse the performance of mechanical homogenizers used in the commercialization of grains in Brazil, as a function of the size of the working sample masses during grain classification. The samples were homogenized and diluted in Boerner, 16:1 multichannel splitter, and 4:1 multichannel splitter until reaching masses of 0.025, 0.050, 0.075, 0.100 and 0.125 kg to determine the level of damaged grains. A 3 x 4 x 5 factorial design was used, meaning three treatments relative to homogenizers (Boerner, 16:1 multichannel splitter, and 4:1 multichannel splitter), four dilutions (4, 8, 12 and 16% damaged grains), and five grain sample sizes (0.025, 0.050, 0.075, 0.100 and 0.125 kg) with nine repetitions. The means were compared by Tukey test and to the original means of prepared samples (4, 8, 12, and 16%) by Student’s t-test. Working samples can be utilized with masses between 0.025 and 0.125 kg to classify damaged soybeans grains. The devices Boerner, 16:1 multichannel splitter, and 4:1 multichannel splitter are similar in the reduction and homogenization of soybean samples for different levels of damaged grains and sample sizes.