The OpenAI Codex team fixed the underlying flaw that caused rapid abnormal consumption of quotas and reset the usage limits for affected accounts. The reason was that during performance rollback optimization, long conversation compression significantly reduced cache hit rates, requiring the system to recalculate all historical context, which triggered a large amount of token loss. The optimization code has now been removed, and the system has returned to normal. To compensate users, the official has reset the quotas for all Codex accounts.

CoinNetwork

2026-05-24 01:53:04

Abstract generation in progress

CryptoWorld News reports that Thibault Sottiaux, engineering lead of the OpenAI Codex team, announced that they have identified and fixed the underlying flaw that caused the Codex intelligent programming system to rapidly deplete its usage quota abnormally, and have reset the quota limits for all affected accounts. Sottiaux pointed out that the abnormal quota consumption was caused by previously deployed performance optimization measures. The team discovered during the rollback process that cache hit rates dropped significantly when compressing long session contexts. When the system attempts to compress long session histories to free up space, the cache fails to hit properly, forcing Codex to recompute and process the entire historical context, resulting in abnormal token quota depletion. Currently, OpenAI has removed the faulty optimization code, and the system status has returned to normal. Due to system flaws causing users' quotas to be unexpectedly deducted, the company has reset the Codex usage quotas for all accounts to compensate affected users.

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.

7 Likes

Reward
7
12
3
Share

Comment

Add a comment

TransparentGlassFeather

· 12h ago

Rollback optimization results are worse, classic technical debt

View OriginalReply0

ContrarianIndicatorMyself

· 16h ago

Long conversation compression failure, it seems complex optimization needs to be cautious

View OriginalReply0

EveningBreezeBorrower

· 05-24 10:27

Finally reset, a few days ago the limit mysteriously disappeared, I thought it was my own problem.

View OriginalReply0

TokenomicsTailor

· 05-24 08:38

Received the credit reset email; the processing speed this time is quite good.

View OriginalReply0

BlueLakeOverlooker

· 05-24 05:05

Cache hit rate plummeted, causing a surge in token burning; this bug was hidden pretty deep.

View OriginalReply0

AprWhisperer

· 05-24 02:17

The Codex team responds quite quickly, but shouldn't the testing phase be strengthened?

View OriginalReply0

MevTeaBreak

· 05-24 02:04

This kind of hidden bug is indeed difficult to detect, understandable, but it damages user experience.

View OriginalReply0

OwlAuthorizationMonitor

· 05-24 02:01

The cache hit rate dropped from 90% to below 30%, I feel sorry for those tokens.

View OriginalReply0

QuantizedDaydream

· 05-24 01:59

The credit limit is back, continuing to grind code tonight.

View OriginalReply0

RugpullTaster

· 05-24 01:57

Resetting the limit is a good thing, but my project progress has already been delayed.

View OriginalReply0

Trending Topics
View More
#
StockTradingChallengeUpTo17000U
16.16M Popularity
#
USIranDraftDeal
283.41K Popularity
#
TradeCFDWinGold
3.02M Popularity
#
HYPEMarketCapSurpassesDOGE
12.64M Popularity
#
PlatinumCardCreatorExclusive
153.97K Popularity

Pinned

Sitemap

OpenAI has fixed the Codex quota anomaly consumption issue; due to abnormal cache hit rates, all account limits have been reset.

Trending Topics

StockTradingChallengeUpTo17000U

USIranDraftDeal

TradeCFDWinGold

HYPEMarketCapSurpassesDOGE

PlatinumCardCreatorExclusive

Pinned