> Instead of a single "one batch with maybe up to 24hr latency" the offer could be a series of tiered queues with different SLAs and costs The API includes a `

> Instead of a single "one batch with maybe up to 24hr latency" the offer could be a series of tiered queues with different SLAs and costs The API includes a `completion_window` parameter ("The time frame within which the batch should be processed. Currently only 24h is supported."). https://platform.openai.com/docs/api-reference/batch/create https://platform.openai.com/docs/api-reference/batch/create