OpenAI Developers Adds Discounted 2 Hour and 8 Hour Response Windows for GPT 6.2 Astra and Sol
Not every workflow requires an immediate response. For single- or multi-turn workflows that can wait, save up to 40% per token than Standard on GPT-6.2 Astra or GPT-6.2 Sol. In Responses API background mode, choose a 2h or 8h target for the first response. These discounted windows are now in public beta.
Context
Letting developers trade speed for price on its top models makes overnight agent work and batch style jobs cheaper, a practical change for anyone running large workloads on the API.
OpenAI Developers posted that single or multi turn workflows that can wait can save up to 40% per token compared with Standard on GPT 6.2 Astra or GPT 6.2 Sol. In Responses API background mode, developers choose a 2 hour or 8 hour target for the first response. OpenAI said the discounted windows are in public beta, and a follow up post gave an example of sending meeting notes before logging off and asking for a morning brief.
Deep dive
The OpenAI Developers account posted that not every workflow requires an immediate response. For single or multi turn workflows that can wait, it said developers can save up to 40% per token compared with Standard pricing on GPT 6.2 Astra or GPT 6.2 Sol. In Responses API background mode, developers choose a 2 hour or 8 hour target for the first response. The post says these discounted windows are now in public beta.
A follow up post from the same account gave an example: send your meeting notes before you log off and ask for a morning brief. It said setting the completion window parameter to 8h gives the API an eight hour window for the first response.
The feature is a pricing and scheduling option rather than a new model. It lets developers accept slower first responses in exchange for a lower price on the same models, which suits overnight summaries, long agent runs, and other jobs where nobody is waiting on the answer. The posts do not give exact per token prices for each tier, so this story does not list them.
The launch comes as GPT 6.2 Astra and Sol are being tested and discussed widely, including early tester reviews and fixes noted on this deck by OpenAI's Tibo Sottiaux and by Theo.
What to expect next
OpenAI's pricing page and API docs should show the exact discounted rates for each window and model. Watch for developers reporting how often responses actually arrive well before the target, and whether the option extends to other models. Competing labs may add similar slow but cheap tiers. If OpenAI later brings a similar option to ChatGPT itself, as some users asked, that would be a separate story.
