Model selection? Nah, you just get to use “GPT-6” or 5.6 Sol. And this generic “GPT-6” option has been noticeably worse than GPT-6 Astra was when I could actually select Astra as the model in Chat…why has explicit model selection been removed?
Reasoning? Nah, you get a really quick initial response regardless of reasoning level, even when the model isn’t done reasoning. And this, in my experience, frequently leads to an incorrect initial answer followed by corrections or contradictions in the same response, making answers unnecessarily confusing and difficult to follow. We need to be allowed to disable premature output and allow the model to finish its reasoning before providing a completed response.
“Pro” reasoning effort? You’re not actually selecting a reasoning effort, you’re just routing it to GPT-6 Astra, and you don’t get to pick the reasoning effort that Astra uses. Why are reasoning effort and model selection being conflated in the same control? Astra needs to be independently selectable in Chat with it’s own reasoning levels.
In the end, I’m fine with OpenAI simplifying the default user experience for users who’d get confused by having a bunch of different options available, but that shouldn’t come at the expense of advanced users who want precise control of the models they are paying to use.
This could easily be done with a “Advanced User” toggle buried in the settings menu that, when checked, allows explicit control over model selection, reasoning effort, and the ‘intermediate response’ feature.