In the ever-expanding and collaborative vehicular network, the formation of vehicle clusters is crucial for maintaining stable communications within each cluster. However, traditional cluster generation schemes often prioritize prolonged maintenance time rather than the better collaboration performance. This is due to the inherent conflict between extended maintenance and optimal collaboration. To tackle this issue, a novel cluster generation scheme is introduced with a specific focus on collaboration, particularly task offloading. The scheme jointly optimizes task offloading and cluster generating decisions to minimize the total task execution delay, while considering the cluster stability issue. Moreover, to ensure efficient resource utilization for vehicles with limited resources, the scheme jointly optimizes the allocation of computing resource and transmission power. To solve the optimization problem, an alternating iterative algorithm with fast convergence is designed. Numerical results demonstrate that, compared with the baseline schemes, the proposed scheme can achieve the minimum total task execution delay and enhance cluster stability.