OpenAI Adds 2- and 8-Hour Completion Windows With Token Discounts
The Responses API feature is in public beta on four models. Discounted requests require Standard processing and cannot stream responses.

OpenAI has put two- and eight-hour completion windows for Responses API background tasks in public beta, giving developers lower token rates when requests can wait longer to finish.
The two-hour window costs 30% less per token than Standard processing, while the eight-hour option costs 40% less. The default “soonest” completion window has no discount. The beta is available on GPT-6.2 Astra, GPT-6.2 Sol, GPT-6.1 Terra and GPT-6 Luna.
Requests using the discounted windows must use Standard processing and cannot stream responses. They also share rate limits with the Batch API.
OpenAI is ending Flex processing for new models beginning with GPT-6.2 Astra and GPT-6.2 Sol. For follow-up requests, reusing the same continuation ID can provide higher queue priority.