Abstract Background Oral prednisone has been recognized as the first-line therapy for the treatment of ocular myasthenia gravis (OMG). However, its long-term use is complicated by numerous adverse effects and is ineffective for some OMG patients in reaching remission. This study aimed to evaluate the effectiveness and safety of intravenous methylprednisolone (IVMP) and tacrolimus monotherapy for OMG patients with unsatisfactory responses to conventional prednisone therapy. Methods We retrospectively reviewed 57 OMG patients who had not achieved satisfactory improvement after prednisone therapy and thereby received IVMP or tacrolimus monotherapy for at least 6 months. Ocular symptoms were evaluated by the ocular-quantitative MG (QMG) score at each time point. A ≥ 2-point fall in ocular QMG score was defined as the cut-off point to indicate clinical improvement. Logistic regression analysis was performed to identify factors associated with the efficacy of IVMP at discharge. Adverse events were recorded. Results Both IVMP and tacrolimus monotherapy demonstrated significant clinical efficacy, with no statistical differences observed at the study endpoint. The proportions of patients who reached the cut-off point for efficacy evaluation were higher in the IVMP group than in the tacrolimus group (1, 3, and 6 months: 51.7% (15/29) vs 12.0% (3/25), p = 0.002; 69.0% (20/29) vs 40.0% (10/25), p = 0.033; 69.0% (20/29) vs 46.4% (13/28), p = 0.085, respectively). Multivariate logistics analysis showed that high ocular QMG scores at baseline indicated favourable responses to IVMP treatment (OR = 1.781; 95% CI 1.066–2.975; p = 0.028). All the adverse events were transient and tolerable. Conclusion Our findings suggest that both IVMP and tacrolimus monotherapy hold promise as viable treatment options for OMG patients with unsatisfactory responses to oral prednisone. The study supports the safety and effectiveness of both therapies, with IVMP exhibiting faster improvement and favourable efficacy in patients with high ocular QMG scores.