Abstract: In order to identify the sweet and bitter identification indicators of almonds, and provide theoretical basis for almond breeding and deep processing. 88 apricot kernels from southern Xinjiang were used as experimental materials. The content of amygdalin and sugar composition were determined by high performance liquid chromatography and gas chromatography mass spectrometry. The results showed that amygdalin was contained in all apricot kernels, and the content ranged from 0.945 mg/g to 36.055 mg/g, the average was 6.236 mg/g and the variation was 118.289%. 69 samples were sweet kernel. Meanwhile the amygdalin content were detected ranging from 0.945 mg/g to 12.148 mg/g, and the amygdalin of 19 samples were detected 10.205 mg/g to 36.055 mg/g. A total of 7 kinds types sugar were detected and sucrose content was the highest, followed by raffinose, and sorbitol content was the lowest. The sweet/bitter kernels was related to content of amygdalin and glucose. The total sugars, sweetness value, sucrose, amygdalin and other factors could affect sweetness/bitterness kernels by principal component analysis. 88 samples were clustered into class Ⅳ by cluster analysis. The class I was extremely high amygbin content, the class II was high total sugar, sucrose, raffinose and sweetness value, the class III was medium amygbin, total sugar and sucrose content, while class IV was low total sugar, sucrose and raffinose content. In conclusion, 'Saimaiti', 'Tuohutikudou' and 'Da huang xing' with high total sugar content, low amygdalin and sweet taste were selected for apricot kernels processing.