It is widely accepted that m6A exhibits significant intercellular specificity, which poses challenges for its detection using existing m6A quantitative methods. In this study, we introduce Scm6A, a machine learning-based approach for single-cell m6A quantification. Scm6A leverages input features derived from the expression levels of m6A trans regulators and cis sequence features, and found that Scm6A offers remarkable prediction efficiency and reliability. To further validate the robustness and precision of Scm6A, we applied a winscore-based m6A calculation method to conduct m6A-seq analysis on CD4+ and CD8+ T-cells isolated through magnetic-activated cell sorting (MACS). Subsequently, we employed Scm6A for analysis on the same samples. Notably, the m6A levels calculated by Scm6A exhibited a significant positive correlation with m6A quantified through m6A-seq in different cells isolated by MACS, providing compelling evidence for Scm6A9s reliability. Additionally, we performed single-cell level m6A analysis on lung cancer tissues as well as blood samples from COVID-19 patients, and demonstrated the landscape and regulatory mechanisms of m6A in different T-cell subtypes from these diseases. In summary, our work has yielded a novel, dependable, and accurate method for single-cell m6A detection. We are confident that Scm6A have broad applications in the realm of m6A-related research.