The peripheral blood cells assist an individual's immune system and protect against viruses that cause illness. Effective classification of these types of cells is essential for medical diagnosis because they contain valuable information that can be examined to establish a person's health status. Since it provides important information for determining the presence of disorders, including viruses and tumors, accurate blood cell type classification is essential to histopathology. In this study, we used DL method for blood cells classification using EfficientNet models (B0-B7) on two datasets Raabin and PBC. The balanced and imbalanced datasets is used in this model to measure his efficiency. The proposed method results shows that EfficientNet B0 perform well on the imbalanced Raabin dataset with an accuracy of 95.8%, and EfficientNet B3 achieves 93.9% accuracy on PBC dataset. In balanced dataset, EfficientNet B3 performs well with 95.4% on Raabin dataset and EfficientNet B0 performs well on PBC dataset with 92.6%. Furthermore, the models are assessed using the evaluation metrics such as, accuracy, recall, precision, and F1 scores. The comparative analysis highlights the adaptability and strength of the models in handling both balanced and imbalanced datasets. These findings indicate the potential of models for improving diagnostic accuracy in medical applications.