Background Evaluating the prognosis of patients with aneurysmal subarachnoid hemorrhage (aSAH) who may be at risk of poor outcomes using grading systems is one way to make a better decision on treatment for these patients. This study aimed to compare the accuracy of the modified World Federation of Neurosurgical Societies (WFNS), WFNS, and Hunt and Hess (H&H) Grading Scales in predicting the outcomes of patients with aSAH. Methods From August 2019 to June 2021, we conducted a multicenter prospective cohort study on adult patients with aSAH in three central hospitals in Hanoi, Vietnam. The primary outcome was the 90-day poor outcome, measured by a score of 4 (moderately severe disability) to 6 (death) on the modified Rankin Scale (mRS). We calculated the areas under the receiver operator characteristic (ROC) curve (AUROCs) to determine how well the grading scales could predict patient prognosis upon admission. We also used ROC curve analysis to find the best cut-off value for each scale. We compared AUROCs using Z-statistics and compared 90-day mean mRS scores among intergrades using the pairwise multiple-comparison test. Finally, we used logistic regression to identify factors associated with the 90-day poor outcome. Results Of 415 patients, 32% had a 90-day poor outcome. The modified WFNS (AUROC: 0.839 [95% confidence interval, CI: 0.795–0.883]; cut-off value≥2.50; P_AUROC<0.001), WFNS (AUROC: 0.837 [95% CI: 0.793–0.881]; cut-off value≥3.5; P_AUROC<0.001), and H&H scales (AUROC: 0.836 [95% CI: 0.791–0.881]; cut-off value≥3.5; P_AUROC<0.001) were all good at predicting patient prognosis on day 90^th after ictus. However, there were no significant differences between the AUROCs of these scales. Only grades IV and V of the modified WFNS (3.75 [standard deviation, SD: 2.46] vs 5.24 [SD: 1.68], p = 0.026, respectively), WFNS (3.75 [SD: 2.46] vs 5.24 [SD: 1.68], p = 0.026, respectively), and H&H scales (2.96 [SD: 2.60] vs 4.97 [SD: 1.87], p<0.001, respectively) showed a significant difference in the 90-day mean mRS scores. In multivariable models, with the same set of confounding variables, the modified WFNS grade of III to V (adjusted odds ratio, AOR: 9.090; 95% CI: 3.494–23.648; P<0.001) was more strongly associated with the increased risk of the 90-day poor outcome compared to the WFNS grade of IV to V (AOR: 6.383; 95% CI: 2.661–15.310; P<0.001) or the H&H grade of IV to V (AOR: 6.146; 95% CI: 2.584–14.620; P<0.001). Conclusions In this study, the modified WFNS, WFNS, and H&H scales all had good discriminatory abilities for the prognosis of patients with aSAH. Because of the better effect size in predicting poor outcomes, the modified WFNS scale seems preferable to the WFNS and H&H scales.