Prunasin and amygdalin are important factors that influence the kernel taste of apricots, however, the regulatory mechanisms underlying this are unclear. In this study, we analyzed the phenotype and transcriptome of kernels during development in Prunus sibirica (bitter kernels) and Prunus armeniaca × Prunus sibirica (kernel consumption apricot, sweet kernels). Prunasin and amygdalin content was significantly higher in bitter kernels compared with that in sweet kernels. Prunasin content exhibited a decreasing trend in both bitter and sweet kernels. The fastest decline was observed in bitter and sweet kernels during S3–S4 (82.21%) and S2–S3 (59.65%), respectively. The amygdalin content in the bitter kernels exhibited the fastest increase between 45–60 d after flowering, and reached a peak at 6.22% on 60 d after flowering. In contrast, the peak in sweet kernels occurred at 60 d after flowering, with a much lower content of 0.18%. Transcriptome analysis revealed 6,942 differentially expressed genes (DEGs), with a subset of 38 DEGs specifically enriched in the cyanoamino acid metabolic pathway. Among these, the ten candidate genes, including CYP79, CYP71, UGT1, AH, and PH, were identified as crucial in regulating prunasin and amygdalin metabolism. Furthermore, a weighted gene co-expression network analysis (WGCNA) unveiled two modules that exhibited significant correlation with prunasin and amygdalin content. Five DEGs were located at the center of the co-expression network, and were identified as hub genes, with four positively regulating prunasin content and one negatively regulating amygdalin content. Our results provide novel insights into the molecular-level regulation of the apricot kernel taste.