<p>Cloud computing is a technology that delivers storage, processing, and networking capabilities over a network of Internet-connected servers. The technique enables on-demand access to and use of computing capabilities, without requiring on-site pre-installed infrastructure. The providers offer a range of Virtual Machine (VM) instances with diverse configurations to meet users' task requirements. Efficient task scheduling across these VMs poses a significant challenge, affecting overall expenses, processing time, and resource utilization. The task scheduling issue is complex and NP-hard. This study introduces a new approach, the Reptile Search Algorithm (RSA), for scheduling tasks in cloud computing. Though RSA delivers satisfactory results on various optimization problems, its performance suffers from low diversity, unbalanced exploration and exploitation of the best-so-far solution, and local-optimum traps. To address these issues with the RSA algorithm, this study proposes an innovative enhancement approach based on the Adaptive Chaotic Reverse Learning (ACRL) technique. Our approach fully leverages the nature of the chaotic map and can boost the diversity of the RSA algorithm’s population. By applying the ACRL approach to the proposed RSA-based algorithm, we aim to avoid early convergence and promote extensive exploration of the search domain. To analyze the efficiency of the proposed approach in a cloud computing setup, we evaluate its performance using the CloudSim toolkit. Experiment results confirm the proposed approach’s efficiency, with significantly lower makespan, computational costs, and power consumption than those of other algorithms.</p>