<p>Asphalt mixture performance is significantly influenced by aggregate gradation. This study concentrated on optimizing the gradation of ATB-25 asphalt mixtures by integrating an orthogonal experimental design with a multi-indicator evaluation method based on the extension goodness algorithm. Coarse and fine aggregates were treated as primary factors, with three grading curve positions serving as levels, resulting in nine gradation schemes within the specification range. These mixtures were tested under optimal asphalt-aggregate ratio conditions for key performance indicators: dynamic stability, Marshall stability, voids filled with asphalt, water residual stability, freeze–thaw splitting strength ratio, and void ratio. A combined weighting method, incorporating the analytic hierarchy process (AHP) and entropy weighting, was employed to assign weights to each indicator, and the correlation degrees with the ideal evaluation grade were computed. The results demonstrated that the 4# gradation achieved the highest comprehensive correlation degree (0.926 with Grade I), with dynamic stability of 4876 passes/mm and residual stability of 92.4%. However, to ensure the robustness of these findings, statistical analysis was conducted, including the calculation of confidence intervals (± 147 passes/mm for dynamic stability, ± 2.3% for residual stability) and p-values (p = 0.03 for dynamic stability, p = 0.04 for residual stability) for the key performance indices. The reported values for 4# gradation were statistically significant at a significance level of 0.05, indicating superior performance across all indices. The proposed extension goodness algorithm-based method effectively consolidated multi-indicator assessments into a single-objective framework, enhancing both the scientific rigor and practical applicability of asphalt mixture gradation optimization.</p>