With　the　rapid　development　of　the　Internet　of　Vehicles,　vehicles　need　to　complete　lots　of　computing　tasks　in　time.　However,　there　have　been　more　strict　demands　for　latency　due　to　limited　resources　on　board.　In　this　paper,　we　propose　an　offloading　strategy　for　vehicular　tasks　to　minimize　the　maximum　delay　in　the　Vehicle-MEC-Cloud　architecture.　Firstly,　considering　the　urgency　level　of　tasks,　a　novel　price　competition　model　is　constructed　based　on　the　queuing　model　and　preemptive　scheduling　algorithm.　Then,　in　the　Vehicle-MEC-Cloud　architecture,　we　fully　utilize　the　limited　resources　of　each　vehicle　itself　and　formulate　the　optimization　problem.　Finally,　we　utilize　a　variational　inequality　to　convert　the　non-convex　problem　as　a　convex　one,　and　propose　a　optimization　algorithm　based　on　the　combination　of　gradient　projection　shrinkage　algorithm　and　greedy　algorithm　to　solve　the　problem.　The　simulation　results　demonstrate　that　the　proposed　strategy　outperforms　other　existing　strategies　significantly　in　terms　of　delay　and　offloading　ratio.　©　2022　IEEE.