2509 Background: ICI-induced autoimmune diabetes (ICI-DM) is an immune-related adverse event (IrAE) that affects approximately 1% of patients who receive ICI. Most reports of ICI-DM suggest that it resembles fulminant type 1 diabetes, with low or absent endogenous insulin production at presentation. However, milder cases have recently been reported, suggesting a more heterogeneous disease. In this study, we identified distinct clinical phenotypes of ICI-DM that may improve individualized approaches to management. Methods: We performed a retrospective chart review of 16,582 patients treated with either anti-PD1, anti- PDL1, anti-CTLA4, or combination between 2010 and 2022 in a multi-centered academic hospital system. We identified patients with ICI-DM as those with a new diagnosis of type 1 diabetes or drug-induced diabetes and a prescription for medium- or long-acting insulin following ICI therapy; cases were confirmed by chart review by two board-certified endocrinologists. To phenotype ICI-DM based on endogenous insulin production, we reviewed all patients with ICI-DM who had a C-peptide level in our system, which allowed us to identify and subdivide 57 cases of ICI-DM. These patients were further characterized at both initial presentation and up to 30 months of follow-up. Results: We identified six distinct groups of ICI-DM based on the presence (A+) or absence (A-) of diabetes-specific autoantibodies, further divided by level of endogenous insulin production, preserved (B+), low (B+/-) or undetectable (B-). In our cohort, A+B- patients showed to have the lowest time to diagnosis (median 45 days) and high rates of both DKA and hospitalization (67%) at presentation. At the other end of the spectrum, A-B+ patients had a longer time to diagnosis (median 176 days) and no evidence of DKA or hospitalization at presentation. Surprisingly, all groups had similar insulin requirements without significant dosage variations at 30-month follow up. Elevated lipase levels varied widely across groups with the highest rate (66.7%) reported in the A+B- cohort. The prevalence of thyroiditis in our cohort was highest in the A-B- group but did not seem to follow any specific distribution. Conclusions: Our study examined the largest data set to date of deep clinical phenotyping of patients with ICI-DM. We define at least three distinct subsets of ICI-DM, with A+B- patients being most similar to the initially described cases of fulminant type 1 diabetes. Patients with negative antibodies seem to have a milder form of the disease. Our findings also suggest significant heterogeneity and distinct underlying mechanisms amongst cases of ICI-DM. Further studies will be needed to elucidate whether these findings could extrapolate to other individual IrAEs, which may also encompass distinct phenotypes with distinct mechanisms.