The optimization of model parameters is a critical challenge in artificial intelligence (AI), profoundly impacting model performance and efficiency. Conventional approaches, such as grid search and manual tuning, are often time-consuming and labor-intensive. This study explores the efficacy of bio-inspired optimization algorithms in tuning hyperparameters for image classification models. The research categorizes these algorithms into three groups based on their solution-update mechanisms: sequence-based, vector-based, and map-based approaches. We designed experiments to investigate the impact of varying the number of classification categories on the performance and evaluation of bio-inspired algorithms. The results demonstrate that sequence-based algorithms exhibit better adaptability and higher accuracy in datasets with both minimal and extensive category counts. Map-based algorithms achieved the highest accuracy on the CIFAR-10 dataset. In contrast, vector-based bio-inspired algorithms consistently maintained moderate performance across all datasets. This study contributes a comprehensive evaluation of bio-inspired algorithms for hyperparameter tuning, providing insights into their strengths and limitations.