OpenAI adds 2h and 8h background windows at 30% and 40% off
Requests that can wait two or eight hours now pay less on GPT-6.2 Astra, GPT-6.2 Sol, GPT-6.1 Terra and GPT-6 Luna.
OpenAI added two discounted completion windows to background mode in the Responses API on November 17, 2026, in public beta. A request that can wait up to two hours pays 30% below Standard model token rates, and one that can wait up to eight hours pays 40% below.
What changed
- Set
backgroundtotrueandbackground_options.completion_windowto2hor8h. The default window,soonest, has no discount. - The discount covers
gpt-6.2-astra,gpt-6.2-sol,gpt-6.1-terraandgpt-6-luna. - Discounted windows require Standard processing: set
service_tiertodefault. Withauto, a request qualifies only if the API picks Standard. Streaming is not available on2hor8hrequests. 2hand8hrequests share the Batch/Background rate limit with Batch API jobs, and input tokens count at submission.- Follow-up requests in an agent run also get the discount: reuse the window,
background_options.continuation_id, model and project. Standard requests can't reuse prompt cache entries created by2hor8hrequests.
Rates
Standard rates per million tokens for prompts up to 272K input tokens, from OpenAI's pricing page on November 17, 2026, with the 30% and 40% discounts applied:
| Model | Standard input | Standard output | 2h input | 2h output | 8h input | 8h output |
|---|---|---|---|---|---|---|
gpt-6.2-astra | $8.00 | $40.00 | $5.60 | $28.00 | $4.80 | $24.00 |
gpt-6.2-sol | $2.00 | $10.00 | $1.40 | $7.00 | $1.20 | $6.00 |
gpt-6.1-terra | $1.00 | $5.00 | $0.70 | $3.50 | $0.60 | $3.00 |
gpt-6-luna | $0.10 | $0.50 | $0.07 | $0.35 | $0.06 | $0.30 |
OpenAI's Batch API still costs 50% of Standard on these models. The new windows keep the Responses API request flow, so an agent run can send tool results back in follow-up requests and keep the discount.
What to do
For work that can wait, add the window to a Responses API request:
{
"model": "gpt-6.1-terra",
"background": true,
"service_tier": "default",
"background_options": { "completion_window": "8h" },
"input": "..."
}Use a webhook or poll the response ID to collect the result. OpenAI says the API targets completion within the window after it accepts the request.
On RouterPlus
These discounts are OpenAI prices for requests sent to OpenAI's Responses API. GPT-6 Luna per million tokens from the RouterPlus catalog on November 17, 2026:
| Model | Provider | Input | Output |
|---|---|---|---|
| GPT-6 Luna | OpenAI (direct API) | $0.10 | $0.50 |
| GPT-6 Luna | Inference.net | $0.10 | $0.50 |
See all OpenAI models on RouterPlus.
