Present cancer disease models - typically based on cell cultures and animal models that lack the human tumor microenvironment (TME) - are extremely poor predictors of human disease outcomes. Microscale cancer models that combine the micromanipulation of tissues and fluids offer the exciting possibility of miniaturizing the drug testing workflow, enabling inexpensive, more efficient tests of high clinical biomimicry that maximize the use of scarce human tissue and minimize animal testing. Critically, these microscale models allow for precisely addressing the impact of the structural features of the heterogeneous TME to properly target and understand the contributions of these unique zones to therapeutic response. We have recently developed a precision slicing method that yields large numbers of cuboidal micro-tissues ("cuboids", ∼ (400 µm) 3 ) from a single tumor biopsy. Here we evaluate cuboids from syngeneic mouse tumor models and human tumors, which contain native immune cells, as models for drug and immunotherapy evaluation. We characterize relevant TME parameters, such as their cellular architecture (immune cells and vasculature), cytokine secretion, proteomics profiles, and their response to drug panels in multi-well arrays. Despite the cutting procedure and the time spent in culture (up to 7 days), the cuboids display strong functional responses such as cytokine and drug responses. Overall, our results suggest that cuboids make an excellent model for applications that require the TME, such as immunotherapy drug evaluations, including for clinical trials and personalized oncology approaches.