Abstract Purpose Immune checkpoint inhibitors have shown efficacy in patients with microsatellite instability-high/mismatch repair-deficient (MSI-H/dMMR) gastrointestinal (GI) cancers. However, depth and duration of clinical response is not uniform. We assessed tumor mutation burden (TMB) as a response marker in patients with GI cancers treated with immune checkpoint inhibitors. Methods Detailed clinical and response data were collected from six patients with metastatic MSI-H/dMMR GI cancers treated with immune checkpoint inhibitors. Efficacy was assessed by Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1. Tumors and matched normal tissue were profiled by targeted next generation sequencing (127 gene panel, size 0.8 Mb). Impact of included mutation types, germline filtering methodology and different variant allele frequency thresholds on TMB estimation was assessed. Results Objective radiographic responses were observed in all six patients, and complete response was achieved in two of the six patients. Responses were durable (minimum 25 months). TMB estimates were clearly above the two recently reported cut-offs for metastatic colorectal cancer of 12 or 37 mutations per megabase for five of six patients, respectively, while one patient had borderline TMB elevation. TMB did not show an association with extent and duration of response but was influenced by included mutation types, germline filtering method and variant allele frequency threshold. Conclusion Our case series confirms the clinical benefit of immune checkpoint blockade in patients with metastatic MSI-H/dMMR GI cancers and illustrates the vulnerability of TMB as predictive marker in a subset of patients.