Server-side quota accounting over-reports consumption after 5h reset: Pro 5h limit burned in ~41 min / ~1.35M tokens

Open 💬 29 comments Opened Jun 25, 2026 by hahasasa
💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

What issue are you seeing?

After a 5h usage limit resets, my Pro accounts reach usage_limit_reached again within ~41 minutes, on roughly 1.35M tokens of actual billed usage. The same account earlier the same day needed ~156M tokens inside a full 5h window before the limit triggered. So the same plan/account hits the same limit at 1.35M tokens (41 min) vs 156M tokens (5h) ??a ~115x discrepancy that cannot be explained by real token consumption.

This matches the symptoms in #29955, #29968, and the older reports (#26512, #21966, #28525, #20153): quota drains far faster than actual usage, and a freshly reset 5h limit drops to 0% after a small burst of requests.

What version of the Codex App are you using (From "About Codex" dialog)?

Latest

What subscription do you have?

Pro

What platform is your computer?

Windows / macOS (multiple Pro accounts)

What steps can reproduce the bug?

  1. Wait for a Pro account 5h limit to reset (upstream returns usage_limit_reached with resets_at in the near future).
  2. After the reset time passes, send a small burst of normal requests (no large context, no heavy codebase).
  3. Observe usage_limit_reached again within ~41 minutes.
  4. Inspect the actual billed tokens for that 41-minute window ??it is only ~1.35M tokens across the active Pro accounts, not the multi-hundred-million that the same account consumed earlier before the limit triggered.

What is the expected behavior?

  • 5h limit consumption should be proportional to actual token usage.
  • A Pro 5h limit should not be fully consumed by ~1.35M tokens in 41 minutes when the same account previously sustained ~156M tokens over 5h before hitting the limit.
  • plan_type: "pro" should behave like a Pro tier, not a much smaller (Plus-like) tier.

Additional information

Measured timeline on 2026-06-25 (KST), captured from the upstream error stream:

| Time (KST) | Event | resets_at | resets_in_seconds |
|---|---|---|---|
| 10:48:50 | usage_limit_reached | 11:02:30 | 820 |
| 11:02:30 | (limit resets) | ??| ??|
| 11:43:45 | usage_limit_reached again | 16:04:33 | 15648 |
| 13:30:42 | usage_limit_reached (higher tier / ~29d reset) | 2026-07-24 | 2511238 |

Cycle A (11:02:30 reset ??11:43:45 limit) = 41 minutes, actual billed usage for the active Pro accounts combined:

  • ~1,055,295 input tokens
  • ~4,082 output tokens
  • ~6K output on a second Pro account
  • ~1.35M total tokens across 48 successful requests

For comparison, the same account earlier the same day: ~156M tokens in a 5h rolling window before the limit triggered.

Relevant upstream error payload:

status: 429
error: {
  "type": "usage_limit_reached",
  "message": "The usage limit has been reached",
  "plan_type": "pro",
  "resets_in_seconds": 15648,
  "eligible_promo": null
}

Notes:

  • plan_type is "pro", but the effective limit behaves like a much smaller tier.
  • eligible_promo: null ??no bonus-credit path; this is the standard 5h limit being over-consumed.
  • Idle periods between bursts show no usage on my side, yet the limit is still fully consumed right after a reset.
  • The fact that the same account hit the same limit at 1.35M tokens (41 min) vs 156M tokens (5h) is strong evidence of a server-side quota accounting bug where consumption is massively over-reported after a reset, rather than proportional to actual token usage.

Happy to provide more sanitized logs if it helps engineering reproduce.

View original on GitHub ↗

29 Comments

github-actions[bot] contributor · 25 days ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

  • #29955
  • #29963
  • #28823
  • #29604
  • #28879

Powered by Codex Action

m1337v · 25 days ago

Adding to this: Woke up to quota going down even tho i didn't send a single thing (was sleeping)

<img width="705" height="133" alt="Image" src="https://github.com/user-attachments/assets/abcd22a3-72c1-4f8b-8e21-79473e230bc3" />

abcdshiwo · 25 days ago
Adding to this: Woke up to quota going down even tho i didn't send a single thing (was sleeping) <img alt="Image" width="705" height="133" src="https://private-user-images.githubusercontent.com/20713070/612897126-abcd22a3-72c1-4f8b-8e21-79473e230bc3.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.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.OyeNOXKtPHsJz9CT_vixHypNPUNpCKWhtcDHUhHPyiM">

Okay Okay, then I can probably guess what happened.
I had seen someone got pro5x --> pro 20x, someone got pro20x --> plus, you got token flew away without using, someone got banned for no reason.
So I boldly assume that the problem is that Open AI has mixed up the account limits of everyone

  1. Your account may be displaying someone else's account limit: the actual account owner is using his account normally
  2. Other people's accounts may display your account limit: someone else is using the limit that belongs to your account
  3. Your account has been hacked, but this is generally unlikely to happen
DyncEric · 25 days ago

It's OpenAI's bug.

Apache-0201 · 25 days ago

Me too. My 20xpro plan performs just like a plus

jiangiang · 25 days ago

My Plus finish all Quota with single prompt, context compact after limit reset

abcdshiwo · 25 days ago
My Plus finish all Quota with single prompt, context compact after limit reset

Yes someone's PLUS account limit has been wrongly turn into free subscription

This bug doesn't just happen to you. Many people have encountered this same issue

TWe1v3 · 25 days ago

我chovy 发了一个hello后给我的pro 20x额度干没了

gimlee · 25 days ago

token runs out too fast
my wallet can't handle it.

roomfull · 25 days ago

Same here. Both of my Pro 20x accounts have usage limits that feel almost like the free tier.

CodingSha · 25 days ago

same problem!!! help~

northXD · 25 days ago

same, after 1 prompt ( it worked for roughly 60 seconds) and said ran out of 5hr usage limit.

lain-creator · 25 days ago

Same problem.

liangjiawei1992 · 25 days ago

same

burmjohn · 25 days ago

Same issue here, feedback id: 019ef1eb-2e63-74d1-8870-c91ef6894e9a

vb-openai contributor · 25 days ago

hi everyone - vb from openai here, can you please put your effected session IDs in the comments below. you can get so by running /feedback and share the IDs here 🙏

m1337v · 25 days ago

@vb-openai
019eb5b3-46dc-7490-8b52-244612b7151d
019efe50-47cd-7db2-a207-395a718f9ff2
019ed514-9470-7740-8def-b54c7d922966

Wentong-Liu · 25 days ago

same, today I just upgraded to 20x from 5x but the quota got drained at same speed as 5x

northXD · 25 days ago

@vb-openai 019ef8da-58ef-79a1-882e-33e95bc9908c

after 5 hrs passed, with just one prompt and 10 sec of working it used %100 5 hr usage. unbelievable. weekly also dropped by nearly %30

roomfull · 25 days ago
hi everyone - vb from openai here, can you please put your effected session IDs in the comments below. you can get so by running /feedback and share the IDs here 🙏

My issue has been resolved, but my weekly limit is still gone.

llearn · 25 days ago
hi everyone - vb from openai here, can you please put your effected session IDs in the comments below. you can get so by running /feedback and share the IDs here 🙏

019f00f8-c2aa-79c3-9768-221d80bb37f5

jiangiang · 25 days ago

it seems fixed for me, but man, my weekly quota has gone for no reason

LxzBUG · 25 days ago
大家好 - 我是 OpenAI 的 VB,请问能否在下方评论区留下受影响的会话 ID?您可以通过运行命令获取会话/feedbackID,然后在这里分享 🙏

019f0192-835b-7ce2-b85f-1c1d6417475f

LxzBUG · 25 days ago
> 大家好 - 我是OpenAI的VB,请问能否在下方评论区留下旅行的会话ID?您可以通过运行命令获取会话/feedbackID,然后在这里分享🙏 019f0192-835b-7ce2-b85f-1c1d6417475f

After resetting my pro account yesterday, it seems to have become like my free account, with the credit limit dropping to 1% in just half an hour.

hahasasa · 25 days ago

@vb-openai i'm not sure which thread had burned my usage-limits, cuz it's occured suddenlly but here it's my pro20x account that' on the issue (storypalse@gmail.com)

DuyTranSipher · 24 days ago
hi everyone - vb from openai here, can you please put your effected session IDs in the comments below. you can get so by running /feedback and share the IDs here 🙏

019f01cb-5acb-7923-9615-aa9137e61fd6

CitySunRise · 24 days ago
hi everyone - vb from openai here, can you please put your effected session IDs in the comments below. you can get so by running /feedback and share the IDs here 🙏

019efc63-15c5-7343-a550-054502a355a8

KaanSevinc · 22 days ago

Same here. I have Pro and it worked for eight hours at medium speed. Half of my weekly tokens are gone. 019e3d2e-fe5a-7b60-ad85-2a8ac5c4a09d

TeBrick · 21 days ago

Same for me, exactly 41 minutes after starting a job in the Codex CLI, my Pro 5x account was depleted...