Background: While TMB is very dependent on methodology, tissue TMB-H (≥10 mutations/megabase) may predict benefit with ICIs. Pembrolizumab received tissue-agnostic approval for TMB-H unresectable cancers in 2020, but little is known about TMB as a predictive biomarker in mGI cancers. We hypothesized that tissue TMB will correlate with efficacy of ICIs in mGI cancers. Methods: A retrospective chart review identified patients with mGI cancers who received an anti-PD-(L)1 drug and had known TMB at a single academic center from 2012 to 2020. The association of TMB with objective response rate (ORR), progression-free survival (PFS), and overall survival (OS) was analyzed using the Fisher’s exact and Log-rank tests. Survival curves were generated using the Kaplan-Meier method. Cox proportional hazard and logistic regression models were used to adjust for microsatellite status. Significance was prespecified at 0.05. Results: 83 patients were identified and included. The most common cancer types were colorectal adenocarcinoma (AC, n = 29), esophageal/gastric AC (n = 21) and SCC (n = 4), cholangiocarcinoma (n = 11), anal SCC (n = 7), and pancreas AC (n = 7). Average age was 61, average number of lines of prior systemic therapy for advanced disease was 1.3 (range 0-4), and 37% of patients were treated on a clinical study. All patients received an anti-PD-(L)1 drug; 6%, 2%, and 36% also received ipilimumab, cytotoxic chemotherapy, and other combinations, respectively. Among those with esophageal/gastric cancer, 76% had known PD-L1 CPS (84% ≥1, 63% ≥5, 42% ≥10). TMB was primarily determined by Foundation One CDx (87%). TMB ranged from 0 to 54; n = 22 (27%) were TMB-H (of these, n = 10 were microsatellite instability-high (MSI-H)), and n = 61 were TMB-L ( < 10 mutations/megabase; of these, n = 2 were MSI-H). The prevalence of TMB-H and microsatellite stable (MSS) was 14.4%. TMB-L, compared to TMB-H, was associated with inferior ORR (3.5% vs 55.6%; odds ratio (OR) 0.045; p < 0.001) and PFS (median 12.7 vs 29.3 weeks; hazard ratio (HR) 2.70; p = 0.001), but not OS (HR 1.20; p = 0.60). In patients with MSS disease, TMB-L, compared to TMB-H, was associated with inferior ORR (OR 0.13; p = 0.04) but not PFS (HR 1.76; p = 0.07) or OS (HR 0.89; p = 0.79). In subgroup analyses, ORR was not significantly associated to tumor type in all or MSS patients. TMB as a continuous variable, in patients with MSS disease, was positively correlated with ORR (p = 0.02) and PFS (p = 0.04), but not OS (p = 0.59). Among all patients, PFS and OS data is immature (median follow-up 13 and 31 weeks). Conclusions: In a single center retrospective study of patients with mGI cancers treated with ICIs, TMB-H was associated with improved ORR and PFS compared to TMB-L. In patients with MSS disease, ORR remained significant. PFS and OS data are immature. TMB as a biomarker of efficacy with ICIs in mGI cancers warrants further study to guide clinical use.