Short answer: yes. In Prefect 3.x the cleanest way is to create two deployments for the same flow, each targeting a different work queue (e.g., one “high” and one “low”) within your work pool. Typical setup (works well with your Vertex AI work pool): 1) Create two queues in the same work pool with different priorities

prefect work-queue create high --pool <your-vertex-pool> --priority <higher>
prefect work-queue create low  --pool <your-vertex-pool> --priority <lower>

2) Deploy the same flow twice, each pointing at a different queue

from prefect import flow

@flow
def my_flow():
    ...

# High priority deployment
my_flow.deploy(
    name="my-flow-high",
    work_pool_name="<your-vertex-pool>",
    work_queue_name="high",
)

# Low priority deployment
my_flow.deploy(
    name="my-flow-low",
    work_pool_name="<your-vertex-pool>",
    work_queue_name="low",
)

3) Start workers for that pool (they can poll both queues)

prefect worker start -p <your-vertex-pool> -q high -q low

Notes: - Priority is defined on the work queue; workers will pull from higher-priority queues first within the same pool. - If your high-priority runs also need different infra (e.g., larger machines/GPUs on Vertex AI), use separate deployments with different job variables, or even separate work pools, to change resources alongside priority. - Tags/parameters do not route a run to a different queue; queue selection is set on the deployment. Docs: - Work pools and queues in Prefect 3.x - Concurrency limits (helpful if you also want to cap throughput per queue/pool) If you want, share your pool/queue names and I can sketch the exact commands for your Vertex AI setup.