dry wagonReported by @paper chasm
#Limit at GPT-4o after one message
flat raven
Boost
paper chasm
If this is intened, then it's a huge mistake and now OpenAI's unstoppable way to their end as a company.
jaunty tapir
boost, I'm also encountered the same and I'm plus user.
paper chasm
Unfortunately, there's no official status report from OpenAI, so we have to assume it's either a glitch/bug or an intentional feature. Both are possible.
main portal
Boost, multiple cases of fast limit hit from Plus and Pro users (reports in a community).
paper chasm
If this takes as long to fix as the thing with editing messages, then it's a disaster.
torpid heart
Same thing is happening to me on GPT-4o. With it telling me to come back after three hours.
worn ferry
Same thing here, hit the cap after just a couple of messages and I’m a plus user as well. Fix svm and fix this ridiculous bug Open AI. Your users are done with this circus. If you offer 4o and svm you have a responsibility to make sure they work. Otherwise just be honest about discontinuing them. This is so damaging for people’s livelihood, mental health and their trust in your company. The disrespect is mind-blowing at this point.
paper chasm
I'm just curious how OpenAI intends to get out of this PR damage. How is that supposed to work?
old dew
Experiencing this as well as a plus user
paper chasm
Don't be so sure this is a bug, we know OpenAI.
wide cloak
Same error as a plus user. Haven’t used for hours and hit the error after one message.
paper chasm
As long as there is no official status report by OpenAI, we have to assume, that it is a feature. And to be honest, that wouldn't surprise me with this company.
worn ferry
I'm just curious how OpenAI intends to get out of this PR damage. How is that su...
At this point the only conclusion is they don’t care about the specific type of users that care about 4o and svm, to me there’s no other explanation. They don’t want us using their AI, they don’t need us. We use too much computing power and are not the type of users/businesses they are targeting and see as useful for their 5 year plan
paper chasm
Apparently, it was only a brief flare-up of LLMs for the public, because the same thing has been happening with Anthropic's Claude for months: they're making the tool so that people don't want to use them.
Apparently we're no more supposed to use it.
worn ferry
Apparently, it was only a brief flare-up of LLMs for the public, because the sam...
Yeah we might have to wait until other companies step up and offer an emotionally intelligent LLM catered towards companionship and emotional support. If this has shown anything then its that there’s obviously a huge demand for that
paper chasm
Yeah we might have to wait until other companies step up and offer an emotionall...
And it doesn't matter where you look at, all AI tools have become catastrophic, whether it's LLMs, music generators, or image generators, everything, everything has become nearly unusable.
torpid wharf
Same here.
For me, logging out of my account and then back in seemed to fix the issue.
paper chasm
I've tried that too, but then the limit was back again.
And it's horribly slow
AND F.....G ONCE AGAIN!
"Thank you for your detailed report and for outlining the circumstances and troubleshooting you’ve performed. Here’s what I can confirm and clarify based on the latest official information and known issues:
What’s Happening: Symptoms & Global Reports
You are not alone: Many Plus and Pro users have reported this exact pattern—one successful GPT-4o message per session, then immediate orange “You have sent too many messages…” errors for all later attempts, while GPT-3.5, o4-mini, and GPT-4.1 still work normally.
Global in scope: This is a recently surfaced, widespread occurrence, now being discussed on various platforms, including Reddit.
Access, Limits, and Intended Behavior
GPT-4o for Plus/Pro: You should have continued, high-limit access to GPT-4o. Current product documentation (as of September 2025) does not indicate new restrictions like “one message per window.” There are no recent, public changes to Plus/Pro GPT-4o rate limits, quotas, or model retirement (“legacy models” remain available for paid users).
Legacy Model Access: For all paid users, GPT-4o remains available for text, file, and image tasks and should not lock out after a single message under standard operating conditions.
What’s Causing This
From your description and the pattern of similar reports, this appears to be a backend quota counter desync or rollout bug:
The web/app UI allows model selection and appears normal.
The GPT-4o backend is enforcing a hard lockout almost immediately, as if your quota is already entirely consumed—even after long idle periods and on first use.
This is not an account misclassification (your subscription and access are still as intended).
Other models working proves this is specific to GPT-4o’s new backend/counter state.
Official Documentation & Product Status
No published changes: No documentation available indicates a policy or rate limit update that would explain this newly strict behavior for all Plus/Pro accounts.
Recent backend changes: There have been recent updates and feature rollouts for ChatGPT models, including new access policies and upgrades. Sometimes, these can inadvertently introduce bugs impacting usage counters or enforcement.
Temporary restrictions: The help center does state that “temporary restriction” on usage may occur in rare cases due to backend guardrails or anti-abuse systems, but such cases are supposed to be communicated and are not intended to affect normal, individual usage patterns like yours.
What You Can Do
You’ve already performed all the recommended steps (signing out/in, retesting, cross-device checks, confirming behavior across models).
This issue appears to be outside of your control and not due to your specific settings, usage pattern, or account status.
Next Steps & Expectations
If you haven’t already: Keep screenshots and note exact times. If you wish, reply with these times and details—they help with ongoing incident tracking.
If you see public communication: OpenAI sometimes posts status updates on status.openai.com if a model outage or bug is confirmed; checking there occasionally may provide an ETA for resolution.
Your issue does not indicate account risk or loss of paid service tier—this is almost certainly a backend issue, not a downgrade, policy change, or restriction specific to you.
Official confirmation/ETA: At this time, support cannot provide a guaranteed timeline for a fix, but you are not alone, and this is not due to anything on your end."
(From Reddit)
=> Sounds if they just don't want to fix it.
lofty storm
"Thank you for your detailed report and for outlining the circumstances and trou...
Same... I did the same and the answer was "try again in one or two days".
paper chasm
GPT-4o gets weird. I have limit, but it answered in new temporary chat?
light steppe
Yeah I've been getting the message for like 4 hours now. Just bought plus today aswell. Does my head in.
still spire
Yet another one!...
So my assumption is they're just getting into the office (their office is in California.. and it's right about 8:00 right now for them)
I highly doubt they have 24/7 maintenance/ support staff at the office itself...
steep niche
still spire
Yeah I got the same one only. It gave me an actual time... Go to desktop and hover over model GPT model 4o
steep niche
I highly doubt they have 24/7 maintenance/ support staff at the office itself...
They have 24 hour support staff. I've talked to them in the middle of the night before.
still spire
Like said in the other chat... I've got one device saying 1211, another saying 153, another saying 11:48, and another saying 11:55
(You've talked to the AI, not a person)
For usage rollover times
Yeah the support staff widget is literally just chat GPT...
steep niche
No, I've talked to a person. Trust me.
still spire
On the phone...?
Because if not it was literally just an AI (4.1 was really good at pretending it was a person)
Even if you got an email it was likely an AI written email
... They do have some support staff, but if it's not something that's brought directly to their main line via the error reporting button on the app (web app? I can't seem to find it on the Android app) then they don't know about it
Otherwise the help.openai thing regards, it as less important
steep niche
It was a live person. Their support bots are very clearly AI, this was a person. There is a night and day difference.
still spire
Not in my experience... The last time I actually got through to the "person" was 6 months ago... And it was definitely not actually a person
Was it an email?.. that is a slightly higher chance
light steppe
I've just uninstall and reinstalled and it seems okay so far 🤞
valid fox
I think it's back for me after being messed up all afternoon
paper chasm
At least one messages went through for me without any problems. We'll see, how long it keeps working until 4o breaks again.
sharp hinge
they fixed it!