The Codex usage limits for ChatGPT free and Go plan users switched to monthly resets this week. The old weekly cycle disappeared for a lot of accounts without any heads up from OpenAI.

I checked the threads that started popping up Monday evening. Screenshots showed the usage bar and a reset timer pointing to roughly a month out.

The immediate fallout

Go plan subscribers and free users both reported the same thing. Their Codex access now sits behind a longer wall. One person said they burned through a serious chunk of the new limit on just a few basic prompts using the current flagship model.

Another described running a short project session and watching the whole monthly allowance disappear inside two hours. The weekly version at least reset before frustration peaked for most light users.

HarshithLucky3 posted one of the first clear examples with the updated timer visible. Ritish_1618 flagged the Go subscription change the day prior.

The quota math is still unclear

What nobody has pinned down yet is the actual size of that monthly bucket. Previous weekly limits multiplied across four weeks gave a certain total. The new monthly figure may or may not match that total.

Replies in both threads lean toward it being smaller in practice. People who track their usage closely say the effective daily allowance feels tighter once you account for how quickly tasks consume it now.

(This kind of unannounced shift usually means the company wants to see how much pushback lands before saying anything official.)

The original poster on the bigger thread offered to test a Go account themselves and post results. That kind of user driven verification is filling the gap while OpenAI stays quiet on the exact numbers.

Free accounts already operated under tight constraints. Spreading a limited pool across thirty days instead of seven just concentrates the restriction. Go plan users who pay a lower entry fee landed in the identical situation.

OpenAI's Codex help page still describes limits by plan and notes they count toward overall agentic usage. It does not mention the reset window change that hit free and Go tiers this week.

Their current documentation sits here if you want the baseline language. The lived experience is showing up in the replies instead.

Some users mentioned the 5.3 Codex option got removed earlier, forcing everything through the heavier 5.5 model that eats quota faster. That compounds the monthly reset pain for anyone doing real work.

The backlash is already rolling in the usual places. Until OpenAI clarifies the quota size or adjusts the reset, free and Go users will keep doing their own calculations every time the bar moves.