Background <p>Accurate prediction of pre-blast fragment size distribution is critical, as it directly influences the downstream operations such as loading, hauling, and grinding. This study evaluates a modified Kuz–Ram model and its suitability in tunnel blasting to predict the size distribution of fragments. </p> Methods <p>The prediction of mean fragment size (X₅₀), uniformity index (n), and characteristic size (X_C) are evaluated in three different types of rock masses: dolomite, augen gneiss, and phyllite. The model provides reasonable estimation for dolomite and phyllite but significantly underestimates the fragment size in augen gneiss. </p> Results <p>The predicted values of mean fragment size exhibit a large deviation from actual measurements, with a mean absolute deviation of 18.25 cm and a root mean square deviation of 18.29 cm in augen gneiss. This deviation suggests that the accuracy of the modified Kuz–Ram model decreases with increasing Rock Mass Rating (RMR), Geological Strength Index (GSI), and joint spacing values. Notably, the value of these parameters are higher in the augen gneiss than in the other two rock types. The multiple non-linear regression analysis was employed to refine the model for tunnel blasting. The resulting model significantly reduces the error margins from over 60% to approximately 10%. Similarly, correlation equations between predicted and actual measurements of uniformity index and characteristic size were developed, with a correlation coefficient (R²) of 0.71 or higher. A novel empirical relationship correlating X₅₀ with RMR, GSI, and average joint spacing was established with an R² value of 0.70 or greater for the various rock masses.</p> Conclusion <p>These newly developed equations are appropriate not only for this specific rock type but for other types of rock masses with similar properties. The findings of this study offer a more practical tool for underground blasting.</p>