After Six Days in Codex, I Switched My GPT-6 Luna Worker from XHigh to High

For the last 6 days, I still use gpt-6 luna max as my daily driver, but changed the worker to luna high. It still does the task fine for max to QA later, and it is faster than luna xhigh. The time and cost per task make the switch worth it.

I know it's already cheap enough, but you can get much more work done. luna xhigh still thinks a lot, and it's slow. luna medium looks like it doesn't have a brain, and luna light, idk what that thingy is for, it can't even open a link. But luna high as the worker is fine. Actually, I do use luna med when I have something with skills, but yeah, you don't need to build a skill all the time.

How I see Codex works:

Plus = Luna users,

Pro = Sol users,

and I guess Pro Max = Astra users.

People should stop crying that your usage burns so fast when you use astra for the wrong jobs on Plus. I mean, luna can do a lot of things for you. sol and astra you definitely still need to use them, but they are just specialists here and there.

I see Opus 5.5 looks really good for dev, and people should switch if it does better for them. But for me, with mostly non-hardcore dev stuff, luna is already really good. Not everyone needs to do AI-generated stuff or heavy dev work. Of course, things still need improvement. No one's really happy yet.

The question is, can OpenAI pull the devs back?