Repository navigation
Pro Lite: sudden weekly quota depletion + banked reset and gpt-reserve disappeared without redemption #41969
Description
What version of Codex CLI is running?
codex-cli 0.150.0 (shown in /status after updating the CLI on Sep 1, 2026).
What subscription do you have?
ChatGPT Pro Lite / Pro 5x.
Which model were you using?
gpt-5.6-luna with Max reasoning.
What platform is your computer?
macOS, Apple Silicon.
What issue are you seeing?
I am seeing three account/usage symptoms at the same time:
-
Weekly quota depleted much faster than my established baseline.
- Previous night: approximately 79% remaining.
- Morning of Sep 1: approximately 37% remaining.
- That is a drop of about 42 percentage points overnight.
- With the same Luna Max development workflow over the previous ~2 weeks, an overnight run normally consumed only about 10–12 percentage points.
-
A previously visible banked usage reset disappeared.
- I did not intentionally redeem/apply the reset.
- The current weekly window still shows roughly 37% remaining and a reset around Sep 7, so I do not see evidence that a normal full reset was successfully applied.
- Please verify the server-side reset ledger and report whether the reset was redeemed, expired, revoked, or removed for another reason, including the exact timestamp.
-
The previously visible Luna/GPT Reserve allowance is also absent.
gpt-reserve/ Luna Reserve was visible on this account previously.- It disappeared at the same time as the other usage anomalies.
The Codex/Work Analytics page still shows recent CLI activity, but the amount of useful work performed does not appear consistent with the size of the quota drop compared with my own two-week baseline.
Timeline
- Aug 31 night (IST): ~79% weekly allowance remaining.
- Sep 1 around 08:15 IST:
/statusshowed ~37% remaining. - I updated the Codex CLI shortly before checking the state (roughly 30 minutes before noticing the missing reset/reserve indicators).
- I did not manually redeem the banked reset.
Additional diagnostics
After noticing the issue I inspected running Codex/ChatGPT processes with:
ps -axo pid,etime,command | grep -i '[c]odex'The visible processes were recent app/render/service processes with elapsed times around tens of minutes; I did not find an obvious Codex process that had been running unattended all night.
Current /status also showed the separate GPT-Codex-Spark allowance at 100%, while the normal weekly allowance was ~37% and the Reserve entry was absent.
What steps can reproduce the bug?
This is account/time dependent rather than a single deterministic command, but the observed sequence is:
- Use a Pro Lite account with a banked reset and Luna/GPT Reserve visible.
- Use Luna Max for a normal long-running development workflow that has had a stable consumption baseline for roughly two weeks.
- Observe the weekly quota before leaving the task running (~79% remaining in this case).
- Recheck the next morning.
- Observe a much larger-than-baseline quota drop (~37% remaining), while the previously visible banked reset and
gpt-reserveare also gone. - No reset was deliberately redeemed.
Expected behavior
- Comparable Luna Max workloads on the same account should have broadly comparable quota consumption unless there is a clearly communicated metering/runtime change.
- A banked reset should remain available until it is explicitly redeemed or reaches its disclosed expiration; if removed, the UI should expose the event/reason.
- Luna/GPT Reserve should not silently disappear from an account that previously had it without a clear entitlement explanation.
- Users should be able to audit which foreground/background/model operations consumed each quota bucket.
Requested investigation
Please correlate this account-side event around Sep 1, 2026 ~08:00 IST and check:
- weekly quota accounting and any model/runtime weighting change for Luna Max;
- whether Codex 0.150.x is causing increased model-cycle/context replay or other hidden usage amplification;
- the banked-reset ledger (redeemed / expired / revoked / other state transition, with timestamp);
- the Luna/GPT Reserve entitlement and why it stopped being surfaced;
- any background/agent activity attributed to the account during the overnight window.
I can provide screenshots and sanitized local telemetry/session data if maintainers need them.
Related issues
- [Meta] Abnormal Codex usage/quota depletion and usage-accounting inconsistencies — cross-report tracker #41220 — abnormal Codex usage/quota depletion tracker
- Codex 0.150 processes comparable ~32.8M-token high-reasoning workloads ~1.6–1.8x faster in wall-clock time, rapidly depleting Pro weekly quota #41541 — 0.150.x higher model-step/token throughput on ProLite
- Codex CLI cannot use Luna Reserve after standard usage limit is exhausted #40939 — Luna Reserve/CLI entitlement mismatch
- Banked Codex resets disappeared after a brief subscription lapse and immediate Pro reactivation #39577 — previously granted banked resets disappearing
- "Reset Usage" overwrites healthy weekly quota and destroys gpt-reserve pool instead of extending cycle #41785 / Codex consumed my banked usage reset without explicit confirmation #41801 — recent reset/reserve ledger and reset-authorization failures
Activity
github-actions commentedon Sep 1, 2026
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
romi007r commentedon Sep 1, 2026
Thanks — I reviewed the suggested duplicates. They are clearly related, but I don't think #41969 is an exact duplicate because the account-state symptom here is different.
gpt-reservewere altered.In my case, no banked reset was intentionally redeemed or triggered. The previously visible banked reset simply disappeared, and the previously visible Luna /
gpt-reserveallowance disappeared at roughly the same time, while the normal weekly window still showed about 37% remaining with the next reset around Sep 7. At the same time, weekly usage had fallen from about 79% remaining to 37% overnight, far outside my ~2-week Luna Max baseline of roughly 10–12 percentage points for a comparable overnight run.So this report may share the quota-depletion root cause with #41541/#41957, but it also asks maintainers to check a potentially separate reset/reserve entitlement or ledger transition that occurred without an explicit reset action.
If maintainers determine these are the same backend incident, I'm happy for this to be consolidated. Before closing it as a duplicate, though, could the server-side account ledger please be checked for the Sep 1 ~08:00 IST window to determine whether the banked reset was redeemed, expired, revoked, or otherwise removed, and why
gpt-reservestopped being surfaced?romi007r commentedon Sep 1, 2026
Important clarification on the banked reset: this reset was not merely inferred to exist from prior account history. It was explicitly visible in the usage UI with an expiration date of September 21, 2026 before it disappeared on September 1.
That means normal expiration cannot explain its disappearance. I did not intentionally redeem/apply it, and the weekly window did not show the expected behavior of a successfully applied full reset.
Please check the server-side reset ledger for this specific reset and report whether it was redeemed, revoked, migrated, or otherwise removed, including the exact timestamp and reason. This also makes the simultaneous disappearance of Luna/GPT Reserve more important to correlate with the same account event.
romi007r commentedon Sep 1, 2026
Update / correction on the banked reset:
The banked full reset has now reappeared in the web Codex and Work Analytics → Usage page. It is intact and shows:
So I am no longer claiming that the banked reset was consumed, revoked, or permanently removed. The more accurate symptom is that the reset was temporarily not surfaced in the Usage UI and later reappeared with its original expiry, which suggests a transient UI/account-state synchronization issue.
This correction does not resolve the main usage anomaly:
gpt-reserveallowance is still not being surfaced in the current usage/status views.Please treat the banked-reset symptom as temporary visibility/synchronization, and focus the investigation on (1) the abrupt quota-consumption increase, (2) Luna Max / 0.150.x metering or execution behavior, and (3) why
gpt-reservestopped being surfaced on the account.I have screenshots showing both the earlier state where the reset was not visible and the later web state where the same reset reappeared intact with the Sep 21 expiry.
LimpingNinja commentedon Sep 2, 2026
Confirming that I can see the same from my side, I am on the Pro 20x subscription and August 31st->1st burned through my weekly quota + $50 in backup credit utilization and burned through 7% in 1 hour without changing the model or usage from the preceding week.
It's concerning when OpenAI substantially changes usage like this with no notification during a bill cycle.
UpUpLiu commentedon Sep 3, 2026
确认我这边也看到了同样的情况，我订阅的是 Pro 20x 套餐，8 月 30 日至 31 日期间，我的每周配额用完了，然后我在9月1号早上重置。但是我在 9月3号 基本只剩下了 25% 这还是我将大量任务转为了使用 dsh 的 deepseek flash实现的结果。 我看了我的个人资料，我的token 消耗变的异常的多。
问题是我没有改变使用方式，还是和之前一样。 在8月25号之前，我这样的使用方式，至少能坚持到 每周的 第5天。 而且我9月1号消耗了36个亿的token ， 这基本是我之前1周的用量，这消耗也太快了。 按8月25号之前，9月1号我最多消耗 30%，但是9月1号 当天我就消耗没了。
LimpingNinja commentedon Sep 3, 2026
Adding support for this, I took this sample set from another direction to provide additional support and detail, here's an 11% chunk of Sol Ultra (last week) vs an 11% chunk of Sol Max and Terra Xhigh; the latter (lesser models) provided significantly less wall time and value.
Exact same-band comparison
Window Remaining usage Recorded model / effort Duration Local token-usage updates Raw tokens Model rate-card proxy*
Aug 26 64% → 53% GPT-5.6 Sol / Ultra only 1h 16m 604 229.4M 2,753.3
Sep 2 64% → 53% GPT-5.6 Sol Max/inherited, plus 5h 33m 885 91.3M 1,316.3
33 Terra xhigh updates
For the same 11 percentage points of quota consumption, the Sep 2 interval had:
60% less raw token work: 91.3M vs 229.4M
52% less token-weighted work under the published model rate card: 1,316.3 vs 2,753.3
47% more local token-usage updates: 885 vs 604
73% fewer tokens per update: approximately 103K vs 380K
The Sep 2 workload consisted of many smaller agent/tool-loop requests, not more token-weighted work. The local records show 885 token-usage updates during the interval.
#42448
My system burned the rest of my quota, $50, and 25% of a banked reset in one day.
@UpUpLiu请提交一份包含此信息的支持工单。请将异常使用情况与上周的使用情况进行比较。我以配额的 10% 为例。请参考我的支持工单中的相关信息。或许您可以提供类似的数据，以便他们进行故障排除。参见 #42448 另外，添加到 [meta] 工单 #41220
romi007r commentedon Sep 5, 2026
Any update on gpt luna reserve quota is it permanently removed or only i am impacted by this