The intention was purely making the product experience better, based on common feedback from people (including myself) that wait times were too long. Cost was not a goal here.
If you still want the higher reliability of longer thinking times, that option is not gone. You can manually select Extended (or Heavy, if you're a Pro user). It's the same as at launch (though we did inadvertently drop it last month and restored it yesterday after Tibor and others pointed it out).
Yep, we recently sped up default thinking times in ChatGPT, as now documented in the release notes: https://help.openai.com/en/articles/6825453-chatgpt-release-...
The intention was purely making the product experience better, based on common feedback from people (including myself) that wait times were too long. Cost was not a goal here.
If you still want the higher reliability of longer thinking times, that option is not gone. You can manually select Extended (or Heavy, if you're a Pro user). It's the same as at launch (though we did inadvertently drop it last month and restored it yesterday after Tibor and others pointed it out).
Isn’t that just how many steps at most a reasoning model should do?