The limitations of liver biopsy have led to the development of indirect noninvasive models for liver fibrosis assessment. We aimed to evaluate and compare the performance of 30 noninvasive models to predict fibrosis stage in treatment‐naïve and treated chronic hepatitis B (CHB) patients. A total of 576 Chinese treatment‐naïve CHB patients and 236 treated CHB patients who had undergone percutaneous liver biopsy were included in the analysis. Histological grading and staging was assessed by the Ishak scoring system. The diagnostic accuracies of 30 noninvasive models were assessed by area under the receiver operating characteristic curves (AUROCs). In treatment‐naïve CHB patients, the AUROCs of the 30 noninvasive models for discriminating significant fibrosis (SF) were less than 0.800, and only the AUROC of the PP score for diagnosing advanced fibrosis (AF) was more than 0.800, while the AUROCs of FIB‐4, FibroQ, HB‐F, Lok index, PHP score and PP score for predicting cirrhosis were greater than 0.800. In treated CHB patients, only the AUROCs of APRI, GUCI, King's score and Wang I for identifying cirrhosis were more than 0.800. The Spearman correlation analysis identified that only the changes in FCI and Virahep‐C model values were weakly correlated with changes in Ishak fibrosis scores before and after treatment (r = 0.206, p = 0.008; r = 0.187, p = 0.016, respectively). In conclusion, in Chinese CHB patients, the 30 existing noninvasive models were not suitable for assessing each stage of fibrosis except cirrhosis before and after antiviral therapy, especially in gauging progression and regression of liver fibrosis following therapy.