Skip to content

OpenAI discounts the Responses API for requests that can wait

A new completion_window option trades a two or eight-hour delay for 30% or 40% off model token rates in the Responses API, without the separate job format Batch requires.

Source: OpenAI

OpenAI has added two discounted processing windows to background mode in the Responses API. Setting background_options.completion_window to 2h or 8h cuts the model’s token rate by 30% or 40% against Standard pricing, in exchange for the request taking up to that long to finish instead of running on demand.

The discount works on v1/responses calls for GPT-6.2 Astra, GPT-6.2 Sol, GPT-6.1 Terra and GPT-6 Luna. It sits next to the existing soonest option, which keeps on-demand processing at full price and is the only one of the three that supports streaming. Both discounted windows require stream to be false or unset. The feature shipped in public beta on 17 November.

await client.responses.create({
  model: 'gpt-6.1-terra',
  input: 'Review document for missing approvals',
  background: true,
  background_options: { completion_window: '8h' },
})

Why it matters: this is a lower-friction discount than the Batch API, since it stays inside the Responses API most integrations already call, rather than needing a separate batch job format with its own upload and polling steps. A nightly report, a backlog of document reviews, or bulk classification that doesn’t need an answer immediately can move to an 8-hour window for a real saving with a one-line change.

Share

Related reading

More from The Wire

All briefs