#GPT-4 Usage Cap Halved Around UTC 10-16 for both Plus and Team user

1 messages · Page 1 of 1 (latest)

astral rover
#

Bug Report:

Steps to reproduce:

  1. Wait for the usage limit to fully reset, ensuring the allowance for new message sends.
  2. Send messages consecutively and keep track of the number of messages sent, particularly in the hours around UTC 10-16.
  3. Observe the point at which the usage limit is reached.

Expected result:

Based on the information provided for the specific plan (e.g., ChatGPT Team plan offering 100 messages/3 hours), the expectation was to send the advertised number of messages within a 3-hour window.

Actual result:

Regardless of the advertised limit (e.g., 40 or 100 messages), the usage limit was observed to be halved (e.g., 20 or 50 messages) specifically in the hours around UTC 12, indicating a significant reduction from the expected limits.

Additional information

This test was verified on several different accounts (plus and team) and for at least 10 days.

I don't think this is a bug, it's intentional

frail wing
astral rover
frail wing
astral rover
frail wing
astral rover
frail wing
astral rover
#

😅 It happens again, around 80% (32/3h) for now

astral rover
#

The limits halved again for now (20/3h for plus, 50/3h for team)

obsidian mountain
#

Yep, can confirm this

astral rover
#

GPT-4 Usage Limit Halved Around UTC 10-16 for both Plus and Team user

obsidian mountain
#

I have the same issue! It occurs aroud midnight here in Australia which corresponds to UTC 13 i think

wild horizon
#

same here I just checked and after exacly 20 messegas gpt 4 stops working ... did not use gpt 4 since 9 am this morning stated at 12 again now I need to wait till 15 to get is back what is this ? I mean I pay for 40 every 3h which is still low and I dont even get those ??

obsidian mountain
#

Hopefully openAI devs read this soon

scarlet snow
#

In my case today only with 18 messages I got blocked and I need to wait 2 hours... Does it count as more if you send a longer message? or any retry of totally errored answers when you select that the content is not true or just a useless apology?

eager owl
#

Same for me... And I keep on trying to obtain support with no success...

eager owl
#

I tested today and I can confirm that the problem seems linked to the time of the day. Like some kind of throttle around certain schedules...

astral rover
astral rover
#

The halving of the usage limit happened today at exactly 13 UTC.

obsidian mountain
#

The problem is that it happens suddenly without any warning. If there was a counter I could have planned better.

sour peak
#

i have the same problem! after 20 Messages stopt working! You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again after ...PM.

astral rover
#

It already happened before 10 UTC today

astral rover
#

Day 10 without any official response

#

GPT-4 Usage Cap Halved Around UTC 10-16 for both Plus and Team user

eager owl
#

Imagine if it was the same with your phone... you buy 4 hours, after one month without notice, its lowered to 1 hour, after 3 months you nearly can't use your phone during rush hours... That's what's happening there, at first in November, four DallE images per request, then one, then a limit of 40 requests/3 hours then a floating limit based on an unknown parameter...

graceful trail
#

Wait for the usage limit to fully reset, ensuring the allowance for new message sends.
what do you mean with "fully reset"?

#

I don't think this is a bug, it's intentional
I don't think it is a bug, just misinterpretation on how the limit works

eager owl
graceful trail
#

implying that OpenAI does not know how their systems work sounds more like a anger fueled statement than actually a bug report

#

=/

eager owl
#

That's why we do say that it is intentional and that it is a rule that has been implemented recently without prior warning and/or user information...

graceful trail
obsidian mountain
# graceful trail Discord is not a support channel, despite the fact that some times employees rea...

help.openai.com was not helpful at all so far. I was in contact with support but they suggested the same unhelpful and useless advices like "use a different browser" or "use icognito mode". I tried everything they suggested and told them the effect didnt change. They came up with the same advices again. I told them, I did this already and asked for help again. never heard from them again the last 7 days...

obsidian mountain
graceful onyx
#

Hey folks, had a chance to talk to people internally and this is what I know: we sometimes need to adjust the GPT-4 message cap to avoid exceeding what our servers can handle (at which point bad things happen). Recently we had to lower the message cap during peak hours to keep ChatGPT 4 working properly. We're always working on adding more capacity, so expect the message cap to continue changing. We're going to update the help center article about this, but let me know if you have any suggestions for how to better display this inside ChatGPT.

frail wing
hexed leaf
#

They have altered the max caps during peak hours....pray they do not alter it further.

obsidian mountain
#

They should make it clear the 40 message cap is flexible and changes based on the time of day.

signal harbor
#

I mean the rate limits are different for some reason. I was testing a custom GPT and got a rate limit after sending 10 messages yesterday in ChatGPT web but I could send 20 messages in the app after I got rate limited in the web version!

#

So it's a total of 30 messages in Custom GPTs in the span of 30-40 minutes

#

"Remaining messages" feature below the ChatGPT 3.5 & 4 dropdown can be very helpful. It will make things more predictable. It's really hard to test the custom GPTs with Actions when we don't know how many messages are left for us to use. This problem actually interrupted my work flow many times this week.