CAR T cell therapy for AML remains limited due to the lack of a proper target without on-target off-tumor toxicity. TIM3 is a promising target due to its high expression on AML cells and absence in most normal hematopoietic cells. Previous reports have shown that each CAR component impacts CAR functionality. Here, we optimized TIM-3 targeting CAR T cells for AML therapy. We generated CARs targeting TIM3 with two different non-signaling domains: an IgG2-CH3 spacer with CD28 transmembrane domain (CH3/CD28) and a CD8α spacer with CD8α transmembrane domain (CD8/CD8), and evaluated their characteristics and function. Incorporating the non-signaling CH3/CD28 domain resulted in unstable CAR expression in anti-TIM3 CAR T cells, leading to lower surface CAR expression over time and reduced cytotoxic function compared to anti-TIM3 CARs with the CD8/CD8 domain. Both types of anti-TIM3 CAR T cells transiently exhibited fratricide, which subsided overtime, and both CAR T cells achieved substantial T cell expansion. To further optimize the design, we explored the effects of different costimulatory domains. Compared with CD28 costimulation, 4-1BB and CD27 combined with a CD8/CD8 non-signaling domain showed higher cytokine secretion, superior antitumor activity, and enhanced T-cell persistence after repeated antigen exposure. These findings emphasize the impact of the optimal design of CAR constructs that provide efficient function. In the context of anti-TIM3 CAR T cells, using a CD8α spacer and transmembrane domain with TNFR-based costimulation is a promising CAR design to improve anti-TIM3 CAR T cell function for AML therapy.