Abstract: Threonine, valine, isoleucine, leucine, phenylalanine, histidine, lysine, alanine, arginine, aspartic acid, proline, tyrosine, glycine, glutamic acid, serine and taurine in human colostrum, transitional milk and mature milk were determined by ultra-high performance liquid chromatography after pre-column derivatization with 6-aminoquinolinyl-N-hydroxysuccinimidyl carbamate (AQC), and the determined results were compared with those given by GB 5009.124-2016 (GB method). Thawed breast milk (0.22 mL) was mixed with water, sodium hydroxide solution, hydrochloric acid solution, L-norvaline internal standard solution and hydrochloric acid solution containing 0.1% (volume fraction) phenol, and the mixture was blown by nitrogen for 5 s. After mixing by whirling, the solution was hydrolyzed at 110℃ for (24±0.5) h. An aliquot (0.2 mL) of the hydrolyzed sample was taken from 1 cm below the liquid surface, and mixed with sodium hydroxide solution and hydrochloric acid solution. The mixed solution was passed through a 0.2 μm nylon filter membrane after vortexing, and 10 μL of the above filtrate was derivatized with AQC derivatization reagent for heating at 55℃ for 10 min, and determined according to the optimized working conditions of the instrument. Gradient elution separation was carried out on ACQUITY UPLC BEH C₁₈ column, and detection was performed at the detection wavelength of 260 nm, and internal standard method was used for quantification. It was shown that the linear ranges of the standard curves of the 16 amino acids were 0.5-25 μmol·L^-1, with detection limits (3s) in the range of 2.5-10.7 mg·L^-1. The spiked recovery test was made on the actual sample at the 3 concentration levels, giving results of recovery ranged from 95.0% to 108%, and RSDs (n=6) of the determined values were not more than 9.0%. The proposed method was used for the analysis of breast milk at different lactation stages. Except for histidine in transitional milk and mature milk, the determined values of amino acids in other breast milk were significantly higher than those obtained by GB method (P<0.05), speculating that it was related to the differences in pretreatment method, equipment and sample amount, and the specific reasons needed to be further studied.