CAGE‐B and SAGE‐B scores, consisting of age and fibrotic burden as cirrhosis and/or liver stiffness, were recently proposed to predict hepatocellular carcinoma (HCC) risk among Caucasian chronic hepatitis B (CHB) patients undergoing long‐term antiviral therapy. We externally validated their predictive performances among an independent cohort from Asia, compared to other conventional prediction models. We consecutively recruited CHB patients with well‐controlled viremia (serum HBV DNA < 2000 IU/mL) receiving antiviral therapy. Patients with decompensated cirrhosis or HCC at baseline were excluded. Among 1763 patients, CAGE‐B score provided the highest Heagerty's integrated area under the curve (iAUC) (0.820), followed by SAGE‐B (0.804), mREACH‐B (0.800), CAMD (0.786), mPAGE‐B (0.748) and PAGE‐B (0.721) scores. CAGE‐B score showed a significantly better performance than SAGE‐B, CAMD, PAGE‐B and mPAGE‐B scores, but was similar to mREACH‐B. SAGE‐B score also showed significantly better performance than mPAGE‐B and PAGE‐B, but was similar to CAMD and mREACH‐B. According to CAGE‐B score 0–5, 6–10 and ≥11, the annual HCC incidences were 0.18, 1.34 and 6.03 per 100 person‐years, respectively (all p < 0.001 between each pair). Likewise, by SAGE‐B score 0–5, 6–10 and ≥11, those were 0.31, 1.49 and 8.96 per 100 person‐years, respectively (all p < 0.001 between each pair). Hence, CAGE‐B and SAGE‐B scores showed acceptable predictive performances for Asian CHB patients undergoing antiviral therapy, with the higher performance by CAGE‐B score. They show a trend towards better prognostic capability to predict HCC risk than previous models.