Send background work to a batch tier

You have evaluations, backfills, or summaries that do not need to be interactive.

Tag the request as background QoS

Set the QoS class to background. Zumik routes eligible traffic to a batch tier and reports the outcome.

z.responses.create(
    model="batch.standard",
    input=long_eval_prompt,
    qos={"class": "background"},
)

Once requests flow through Zumik, a diagnostic shows exactly what you are reusing.