Background and aims Baseline National Institutes of Health Stroke Scale (NIHSS) scores have frequently been used for prognostication after ischemic stroke. With the increasing utilization of acute stroke interventions, we aimed to determine whether baseline NIHSS scores are still able to reliably predict post-stroke functional outcome. Methods We retrospectively analyzed prospectively collected data from a high-volume tertiary-care center. We tested strength of association between NIHSS scores at baseline and 24 h with discharge NIHSS using Spearman correlation, and diagnostic accuracy of NIHSS scores in predicting favorable outcome at three months (defined as modified Rankin Scale 0–2) using receiver operating characteristic curve analysis with area under the curve. Results There were 1183 patients in our cohort, with median baseline NIHSS 8 (IQR 3–17), 24-h NIHSS 4 (IQR 1–11), and discharge NIHSS 2 (IQR 1–8). Correlation with discharge NIHSS was r = 0.60 for baseline NIHSS and r = 0.88 for 24-h NIHSS. Of all patients with follow-up data, 425/1037 (41%) had favorable functional outcome at three months. Receiver operating characteristic curve analysis for predicting favorable outcome showed area under the curve 0.698 (95% CI 0.664–0.732) for baseline NIHSS, 0.800 (95% CI 0.772–0.827) for 24-h NIHSS, and 0.819 (95% CI 0.793–0.845) for discharge NIHSS; 24 h and discharge NIHSS maintained robust predictive accuracy for patients receiving mechanical thrombectomy (AUC 0.846, 95% CI 0.798–0.895; AUC 0.873, 95% CI 0.832–0.914, respectively), while accuracy for baseline NIHSS decreased (AUC 0.635, 95% CI 0.566–0.704). Conclusion Baseline NIHSS scores are inferior to 24 h and discharge scores in predicting post-stroke functional outcomes, especially in patients receiving mechanical thrombectomy.