I’m on a plus plan. Weekly usage is typically 5 hours of 5.6 sol high. Today within 90 minutes about 20% of my weekly usage has been used up by just a few Luna Max tasks running in parallel. NOT fast mode. Ridiculous when you look at what im actually supposed to get out of using Luna. Completly pointless to use the model if it still drains like this. Has to be a bug of some kind. Please look into it, i want to have infinite luna usage just like everybody else.
Welcome to the community!
I don’t have any concrete answers for you at this time, I’m just aggregating posts that might be related to your issue.
potentially related (unclear):
churn:
Some questions:
Is it possible that your code-base has grown substantially, such that context usage and context churn might start to become an issue?
There’s also output churn - where the model can’t successfully complete a task and continuously rerolls. Is it possible that that’s something you might be encountering?
They switch any reasoning to Sol, I just asked help this, they didn’t tell anyone as far as I’ve seen.
Neither of these are likely. The task gets completed properly and no matter the context, Luna shouldn’t consume 50% as much usage as Sol. If context would be a problem, it would still be a problem in both models, but the issue is that Luna specifically is draining my usage way more than it should based on API pricing which directly correlated to codex usage.
This has to be wrong. Any reasoning in codex essentially means any anything, because all the models in codex are reasoning models. Switching all reasoning models to.sol seems unlikely, especially since i know people that use Luna with way lower usage drain as it is also intended.
Obviously i have set it on Luna max and didn’t just use the default reasoning. What’s your point?
If Luna max uses as much usage as Sol high that’s on me huh? Yeah right, because the model with 25x lower costs should actually drain my usage just about the same as Sol high. tHaT’s On Me. If you want to ragebait go on tiktok or something, im already pissed off enough by this problem.
Yeah, that’s actually not what it means big dog, that’s why there’s a knob. You think they have you lowest vs max and they cost the same? Longer chains, more time. - if it was the same you wouldn’t start on lowest and go up. Soup yourself a favor, put in on lowest - Does it last longer?
I understand this is hyperbole, but comments about “infinite Luna” are usually made in the context of the $200 Pro plan.
As such, I am not surprised to see usage drop more quickly on a Plus plan.
Without further information, there is not much we can do to help beyond suggesting lower reasoning settings, disabling Fast mode, or reducing context and tool calls, among other options.
I use Pi agent, (sign in with your ChatGPT), and I can have infinite luna max on plus(If you need more than 14 hours a day, I dunno), do not use parallel sub agents for anything, that is just not something you can do on plus. I control my subagents in Pi, limited to one. I could probably do 10 times with luna max than I can for the $20 Claude on Sonnet 5 High, also, update your Codex client! That is something I read on Reddit 2 hours ago regarding high luna max usage.