High-dimensional flow cytometry is a cornerstone of single-cell phenotypic analysis in immunology and oncology. However, the downstream analysis and specifically the manual gating and biological annotation of complex cell populations remains a significant bottleneck. It is time-consuming, susceptible to inter-operator variability, and heavily reliant on domain expertise. While Large Language Models (LLMs) have demonstrated strong zero-shot reasoning capabilities across various biomedical domains, their utility in translating raw immunophenotypic marker profiles into accurate biological cell classifications remains underexplored. To evaluate and compare the accuracy, consistency, and efficiency of three state-of-the-art LLMs in annotating multi-parameter flow cytometry data, we used manual annotations by a panel of domain expert immunologists as the gold standard for both labels and variance. Preliminary results of this comparison will be presented.