On November 17, 2026 OpenAI’s changelog added discounted completion windows to background mode in the Responses API. It is a public beta.
What to know
- How it works: send a background request (
background: true) and setbackground_options.completion_windowto2hor8h. The default,soonest, has no discount. - Discounts: OpenAI says model token rates are 30% below Standard for the two-hour window and 40% below for the eight-hour window.
- Models: GPT-6.2 Astra, GPT-6.2 Sol, GPT-6.1 Terra and GPT-6 Luna.
- Requirements: Standard processing (
service_tier: default), and no streaming for the 2h and 8h windows. - Timing: OpenAI calls the window a target counted from when the request is accepted. Its docs do not say what happens if a response misses it.
- Limits: these requests share the Batch and background rate limit with Batch API jobs, and they cannot reuse prompt cache entries with normal Standard requests.
- Getting results: poll the response by ID, or subscribe to webhooks such as
response.completed. - Data: OpenAI notes background mode stores response data temporarily, encrypted at rest, even with
storeset to false or Zero Data Retention.
Why it matters
Plenty of AI work does not need an answer in seconds: overnight reports, bulk rewrites of product copy, long research runs. Waiting a couple of hours now cuts the token bill without moving to the Batch API’s file workflow. Pair it with the new GPT-6.2 Astra and Sol, or with Terra for screenshot and computer-use tasks.
Related: ChatGPT, Agents API automatic review and our GPT-6.2 prompting guide.
Source: OpenAI API docs. We summarise in our own words; check the source for the full details.
Contribute your creativity