<p>The proliferation of mobile devices has intensified resource constraints on edge servers, revealing significant limitations in traditional mobile edge computing (MEC) for multi-task offloading scenarios. This has created a fundamental tension between stringent task deadlines and inherent system processing delays, which are further exacerbated by queuing dynamics in multi-task MEC environments. Unlike scenarios relying solely on cloud or MEC, the collaborative computing paradigm, where both resources are used in tandem, enables effective complementarity between nodes. This collaboration benefits multi-task offloading by dynamically meeting the demands of real-time task offloading and resource allocation. To address the diversity of task scenarios and the variations in data scale within MEC, and to mitigate the heightened task delay sensitivity resulting from multi-task queuing in MEC, this paper introduces a multi-task offloading cloud-edge collaborative computing system model(C3O). The system dynamically optimizes performance metrics by adjusting delay-payment weight ratios while preserving quality of service. It also incorporates a deadline-based dynamic weight queuing strategy and employs a deep Q-network (DQN) for making offloading decisions. Convex optimization is utilized to balance the weighted sum of payment cost and delay, with this balance serving as the reward signal to guide the agent towards optimal performance. Simulation results demonstrate that C3O is robust across various MEC application scenarios, effectively meeting dynamic resource allocation and delay requirements under multi-task offloading. Compared to other methods, C3O reduces task delay by 7% to 73% and the total task cost by 10% to 78%.</p>