This chapter introduces SRacos, a sequential-mode classification-based derivative-free optimization method designed to address the limitations of batch-mode optimization in scenarios where samples and their evaluations must be obtained sequentially. Unlike batch-mode methods, which require a set of samples to update the model, SRacos updates the sampling model immediately after evaluating each sample by reusing historical data from previous iterations. This approach improves sample efficiency, requiring fewer samples to achieve the same optimization goal compared to batch-mode methods. The chapter provides a theoretical analysis of SRacos, demonstrating its potential for better query complexity under certain conditions. Empirical studies compare SRacos with state-of-the-art optimization algorithms, including CMA-ES, DE, CE, and IMGPO, on synthetic functions and reinforcement learning tasks. Results show that SRacos consistently outperforms batch-mode methods in convergence rate and scalability, particularly in high-dimensional and complex optimization problems. The chapter concludes by highlighting the advantages of sequential-mode optimization in accelerating the optimization process.