Repository navigation
Pro Lite / Pro 5x: weekly quota depleted ~93% in one day after automatic reset despite comparable token volume #42224
Description
What subscription do you have?
ChatGPT Pro Lite / Pro 5x.
What platform / surface are you using?
ChatGPT / Codex Work desktop on Windows.
What issue are you seeing?
My weekly Codex/Work allowance was automatically reset twice over the weekend, and after the second reset the quota depleted dramatically faster than the previous day despite a very similar amount of raw token activity.
The main anomaly is:
- Aug 30: Analytics shows 344.8M raw tokens. I used roughly 30 percentage points of the weekly allowance during the day.
- Aug 31: Analytics shows 331.6M raw tokens, slightly less than Aug 30. After an automatic reset to 100% that morning, the weekly allowance fell to 7% remaining by the evening, i.e. roughly 93 percentage points consumed.
I understand that raw token count is not expected to map 1:1 to quota because cached input, uncached input, output, reasoning effort, model choice, subagents, etc. may be weighted differently. The bug report is therefore not claiming that 331.6M and 344.8M raw tokens must cost exactly the same.
The concern is that a comparable workflow and comparable raw-token volume produced roughly 3x the observed weekly quota depletion immediately after the automatic reset, with no user-visible accounting that explains the discontinuity.
Timeline (Europe/Stockholm)
- Sat Aug 29: approximately 30% weekly allowance remaining. The normal reset was scheduled for Monday.
- Sun Aug 30: allowance unexpectedly returned to 100% without me manually redeeming a reset. The next reset date moved to Sep 6.
- Sun Aug 30 evening: approximately 70% remained after the day's work. Analytics later showed 344.8M raw tokens for Aug 30.
- Mon Aug 31 morning: allowance was again back at 100% without manual redemption. The next reset date moved again, now to Sep 7.
- Mon Aug 31 evening: only 7% remained. The UI showed the next reset as Sep 7, 2026 13:21. Analytics showed 331.6M raw tokens for Aug 31.
For additional context, Analytics shows 66.6M raw tokens on Aug 29.
Secondary reset problem
The automatic resets also replaced the allowance I had deliberately saved before the normal reset date.
For example, if ~30% remained and I was due to receive a normal fresh weekly allowance shortly afterwards, an automatic reset to 100% followed by moving the next reset date a week forward effectively removed the value of the remaining ~30%. Previously, banked resets allowed the user to choose when to redeem them.
I am not assuming this has the same root cause as the abnormal depletion, but it is relevant to the timeline because both quota discontinuities happened immediately around automatic resets.
Expected behavior
- Comparable workloads on the same account should have broadly comparable effective quota consumption unless there is a disclosed model/runtime/metering change.
- If consumption changes sharply, users should be able to audit why.
- An automatic reset should not make already-earned/saved remaining allowance silently disappear without clearly communicating that effect.
- The UI should expose enough per-task accounting to reconcile quota deductions with actual work performed.
Requested investigation
Please correlate the server-side usage ledger for this account around Aug 30–31, 2026 (Europe/Stockholm) and check:
- cached input / uncached input / output / reasoning token weighting;
- model and reasoning-effort changes between the two days;
- Fast/Ultra/subagent/delegated/background activity charged to the same bucket;
- retries, compaction/context replay, polling/wait loops, or other hidden usage amplification;
- delayed/backfilled usage or usage attributed across the automatic-reset boundary;
- duplicate accounting;
- the exact timestamps and reason for the automatic resets on Aug 30 and Aug 31;
- whether quota from before either reset was carried forward, discarded, or otherwise reconciled.
I have screenshots of the Analytics values (344.8M on Aug 30, 331.6M on Aug 31) and the usage-limit UI showing 7% remaining / reset Sep 7 at 13:21, and can provide them 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 and usage-accounting inconsistencies tracker
- Pro Lite: sudden weekly quota depletion + banked reset and gpt-reserve disappeared without redemption #41969 — recent Pro Lite / Pro 5x report of sudden weekly quota depletion after a stable baseline
Activity
github-actions commentedon Sep 2, 2026
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
alexsays1007-cmd commentedon Sep 2, 2026
same on Plus
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:
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.
bompus commentedon Sep 7, 2026
feedback id 01a07a0c-95f6-7da2-a6bf-113be168b169