Purpose: This 2-year longitudinal study examined the unique prediction of a dynamic assessment (DA) of character decoding on Chinese early reading development, and further explored whether learning potential, as assessed by DA, predicts subsequent character reading development in a sample of 135 native Mandarin-speaking Chinese students from first grade to second grade. Method: Latent Growth Modeling was carried out to examine the prediction of DA administered in grade 1 for the character reading growth in Chinese children from grade 1 to 2. Growth Mixture Modeling was used to identify latent-class groups with varying learning potential based on the trial-by-trial progresses on DA, and predicted the distal outcome, Chinese character reading. Results: DA uniquely predicted the final level and growth rate of character reading for Chinese children from grade 1 to 2 after controlling for traditional static predictors. Additionally, three latent groups were identified to characterize different responsiveness to graduated prompts in the DA task; the three latent groups showed significant differences in character reading over time. Conclusion: The findings underscore the importance of DA in predicting Chinese children's early reading development and the classification validity of learning potential subgroups of DA in differentiating early character reading development in Chinese children.