The annotation of cell types on single cell RNA-seq data is a complex, uncertain and time-consuming task, requiring several methods and tools to be able to annotate cells appropriately and efficiently. To overcome these problems and uncertainties, numerous tools and scientific articles have emerged over the years. The rise of artificial intelligence in our lives (notably through chatGPT), has also imposed itself on the scientific world, bringing novelty and innovation to existing techniques and tools. These tools need to be tested and studied to verify their effectiveness. In this project, two cell annotation tools in single cell RNA-seq named scBERT and scGPT are of interest to CB2M because of their ability to resolve and avoid the uncertainties and problems mentioned above. We study here, through various analyses, including cross-validation and the use of multiple qualitative and numerical indicators, that cell annotation by those tools are effective for annotating cells from scRNA-seq. Provided files : Human_Thymus_Development_Atlas_reference.tar.gz : Cell atlas of human thymic development : 15 embryonic and fetal thymuses covering stages of thymic development from 7 post-conceptional weeks (PCW) to 17 PCW, and 9 postnatal thymuses from human pediatric and adult samples. It contains 255,901 cells, 32,922 genes and 33 different cell types. Human_Thymus_Development_Atlas_output.tar.gz : All analysis output files Human_Thymus_Development_Atlas_container.tar.gz : Docker image and Singularity mages used for the analysis See https://github.com/CIML-bioinformatic/CB2M_scTransformers for more details.