AI Catchup

OpenAI Adds Discounted Background Windows to the Responses API

By 2 min read

OpenAI's Responses API now supports 2-hour and 8-hour completion targets for eligible background requests. The public beta discounts model-token rates by 30% or 40% on supported models, including GPT-6.2 Astra and Sol, GPT-6.1 Terra, and GPT-6 Luna.

OpenAI's Responses API now offers discounted completion windows for background requests and multi-turn workflows that can wait for a result. Developers can select a two-hour target for a 30% reduction or an eight-hour target for a 40% reduction from Standard model-token rates on supported models (OpenAI API background mode documentation; OpenAI Developers on X).

The public beta is available for GPT-6.2 Astra, GPT-6.2 Sol, GPT-6.1 Terra, and GPT-6 Luna. The discounts apply to model-token rates on each eligible request, including follow-ups, rather than only to the first turn in a workflow (OpenAI API background mode documentation; OpenAI API pricing).

Choose a completion target

OpenAI's default soonest window remains the on-demand option. The longer windows are for tasks where the application can trade speed for lower model-token rates.

Completion windowDiscount from Standard model-token ratesTarget
soonestNoneThe default window for an on-demand response
2h30%Target completion within two hours after the API accepts the request
8h40%Target completion within eight hours after the API accepts the request

OpenAI says the completion target starts when the API accepts the initial request and applies to that response. For follow-ups, the guide says start times depend on scheduling (OpenAI API background mode documentation).

Configure a discounted request

To use either discounted window, set background to true and set background_options.completion_window to 2h or 8h. The discount requires Standard processing: set service_tier to default, or use auto only when the API selects Standard for the request. Streaming must be disabled or left unset for the two-hour and eight-hour windows (OpenAI API background mode documentation).

OpenAI also documents a continuation option for multi-turn work. Reuse the same project, background setting, background_options.continuation_id, model, service tier, and completion window on follow-up requests to make them eligible for higher queue priority. This raises eligibility for priority, not a stated completion-time guarantee (OpenAI API background mode documentation; OpenAI Developers on X).

Check the operational tradeoffs

Requests using the 2h or 8h windows share the Batch/Background rate limit with Batch API jobs, and OpenAI counts input tokens when the request is submitted. OpenAI lists soonest as the default completion window, and requests using it use the regular request and token limits (OpenAI API background mode documentation).

OpenAI's pricing page lists the discounts as percentages of Standard model-token rates, so teams should consult the current per-model rate card when estimating spend (OpenAI API pricing).

Related coverage

Sources

Frequently Asked Questions

How much can developers save with OpenAI background windows?

OpenAI lists eligible model-token rates at 30% below Standard for a 2-hour completion window and 40% below Standard for an 8-hour window. The discount applies to eligible requests, including follow-ups.

Which models support the discounted completion windows?

OpenAI's public-beta documentation names GPT-6.2 Astra, GPT-6.2 Sol, GPT-6.1 Terra, and GPT-6 Luna.

Is the selected completion time guaranteed?

The documentation describes it as a completion target that starts when the API accepts the initial request; it does not state that completion within the window is guaranteed.

Get the weekly AI Catchup

Tools, practices, and what matters, in your inbox every week.