gpt5-fast-hi used to be worlds apart from gpt5-hi in quality (it should only be priority queue, but it was much more than that).
Is that still the case, or can we now safely use gpt5-hi for half the cost?
Also: Is codex CLI any different (context window..)? For me it’s maybe a bit faster, but else seems same.
Which is a huge compliment on the Cursor devs. But as always, they f*cking leave us in the dark.
Any data appreciated.