Cloud computing is a novel technology that allows consumers to access services from anywhere, at any time, under different conditions, and is controlled by a third-party cloud provider. Cloud task scheduling is a complicated optimisation problem. However, both under- and over-loading conditions cause a range of system problems as far as power consumption, machine failures, and so forth are concerned. consequently, virtual machine (VM) work-load balancing is regarded as a key component of cloud task scheduling. In this paper, a novel cloud-based African vulture particle swarm optimisation [C-AVPSO] has been proposed. Using C-AVPSO, the developed optimization algorithm solves the dynamic load balancing problem effectively. In this method, the exploration space was obtained by using the AVO procedure whereas the enhanced response was identified by the PSO procedure. This algorithm successfully resolves resource utilization, response time, and cost constraints of the task. As a result of combining the AVO and PSO algorithms into the proposed AVPSO algorithm, the convergence rate and performance metrics for load balancing in the cloud environment are improved. To improve the operation's efficiency, the proposed method balances VM loads efficiently. The proposed framework is compared to existing approaches like QMPSO, FIMPSO and ACSO based on energy utilization, degree of imbalance and task migration, response time and resource utilization. The proposed C-AVPSO technique reduces resource utilization of 19.1%, 31%, and 54% than, QMPSO, FIMPSO and ACSO existing techniques.