Repository navigation
Usage allowance is being consumed too quickly #43341
Description
안녕하세요 GPT를 꾸준히 유료해서 쓰고 있습니다
유료해서 쓴지 약 2년이 넘어가는데요
최근 한도 사용량이 과다하게 빨리 끝나거나 합니다.
특히 최근 GPT6 버전 고난도 추론 한 번 사용했더니 끝났습니다.
유료 버전은 [PLUS] 사용자입니다.
5.6SOL도 그럭저럭 빨리 끝나긴 했지만 5.6도 후반돼서 점점 사용량이 빨리 도달합니다;
6은 사용도 못하고 끝나는 느낌입니다
이런 문제를 저만 겪는 게 아니라 대부분 사용자가 겪습니다..
처음으로 GPT에서 다른 생성형 AI로 갈아타야되나 싶을 정도로 사용량이 적습니다..
정식적으로 개선해줬음 합니다..
Activity
github-actions commentedon Sep 7, 2026
English translation:
Hello, I have been using GPT regularly as a paying user.
I've been paying for it for over two years now.
Recently, my usage allowance has been running out far too quickly.
In particular, I recently used high-level reasoning once in the GPT6 version, and that exhausted my allowance.
I am on the paid [PLUS] plan.
5.6SOL also ran out fairly quickly, but even with 5.6, as time went on, I started hitting the usage limit increasingly quickly;
With 6, it feels like the allowance runs out before I can even use it.
I'm not the only one experiencing this problem; most users are experiencing it too..
For the first time, the usage allowance is so low that I'm wondering whether I should switch from GPT to another generative AI..
I hope you will properly address this..
[-]사용 한도 소모율 과다[/-][+]Usage allowance is being consumed too quickly[/+]github-actions commentedon Sep 7, 2026
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
VVNevzorov commentedon Sep 10, 2026
Plus / Windows / VS Code IDE / local / GPT-5.6 Sol High / Standard / fresh conversation / fresh 5-hour window → 100% → 58% after one run.
vostokkostya-spec commentedon Sep 14, 2026
I’m seeing the same problem. I use Codex heavily for development work and have spent approximately $400 in the last month in addition to my subscription, yet my available Codex usage/credits still run out during normal work.
The biggest problem is the lack of transparent accounting. I cannot tell which task, model, context size, compaction, tool call, retry, or agent action caused the quota to drop so quickly.
Please expose per-task usage, input/cached/output token counts, model used, context size before/after compaction, usage attributable to tool calls/retries/sub-agents, and clear warnings before unusually expensive long-running tasks.
At this level of paid usage, I should be able to see exactly where the usage went and why the limits are being exhausted so quickly.