For single-turn or multi-turn GPT-6.2 Sol tasks that do not need an immediate response, you can use background mode in the Responses API and choose a 2-hour or 8-hour window for the initial response, saving up to 40% per token compared with Standard.
Good for: Single-turn or multi-turn workflows that can be delayed, such as submitting meeting notes before signing off and getting a morning briefing the next day.
Not suitable for: Interactions that require an immediate response.
Supported model version: GPT-6.2 Sol.
Supported client, agent, or API: Responses API background mode.
Recommended reasoning effort and parameters: The source does not mention a reasoning effort. completion_window can be set to 2h (30% lower per token) or 8h (40% lower).
Set completion_window to 2h or 8h to specify the target window for the initial response. The source does not provide a complete API request body, so no directly executable request example is added here.
Confirm that the task can wait and does not require an immediate response.
Submit the task in Responses API background mode, choosing 2h or 8h as the initial response window.
If the task is submitted in multiple parts, set a continuation ID when sending the notes and reuse it in subsequent requests to get higher queue priority.
The official post says that waitable single-turn or multi-turn GPT-6.2 Sol workflows can save up to 40% per token compared with Standard. The 2-hour and 8-hour windows are in public beta.
A reply to the same post clarifies that 2h is 30% lower and 8h is 40% lower, and that the selected window applies to the initial response.
A reply to the same post gives the example of sending meeting notes before signing off, asking for a morning briefing, and choosing 8h.
A reply to the same post says to set a continuation ID when sending the notes and reuse it in subsequent requests to increase queue priority. The source does not provide the field name for this ID or the complete configuration format.
The window is a target time range for the initial response. The source does not promise an exact completion time or describe the duration of subsequent responses or the full pricing rules. Public beta availability depends on what is actually enabled in the API.
Access to waitable tasks and Responses API background mode is required. Choose 2h or 8h based on the desired wait time, and reuse the continuation ID in applicable multi-request workflows. The original post does not provide a verifiable request body or execution result.
GPT-6.2 Sol