The discrepancy between estimated glycemia from HbA1c values and actual average glucose (AG) levels has significant implications for treatment decisions and patient understanding. Factors contributing to the gap include red blood cell (RBC) lifespan and glucose uptake into the RBC. Personalized models have been proposed to enhance AG prediction accuracy by considering interpersonal variation. This study contributes to our understanding of personalized models for estimating AG from HbA1c. Utilizing data from seven studies (340 participants), including Hispanic/Latino populations with or at risk of non-insulin-treated type 2 diabetes (T2D), we examined kinetic features across cohorts. Additionally, the study simulated scenarios to understand data requirements for improving accuracy. Personalized approaches improved agreement between AG estimations and CGM-AG, particularly with four or more weeks of training CGM data. A multiple linear regression model using kinetic parameters and added clinical features was shown to improve the accuracy of personalized models further. As CGM usage extends beyond type 1 diabetes, there is growing interest in leveraging CGM data for clinical decision-making. Patient-specific models offer a valuable tool for managing glycemic status in patients with discordant HbA1c and AG values.