Usage went from ~73% to 0% instantly on the $200/mo plan

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

26.901.51231

What subscription do you have?

Pro x20

What platform is your computer?

Darwin 27.0.0 arm64 arm

What issue are you seeing?

Usage went from ~73% to 0% instantly. I was just chatting to ChatGPT on Codex when it suddenly just plunged to 0%.

It was at ~73% immediately prior to this.

What steps can reproduce the bug?

Feedback ID: no-active-thread-01a0870e-6aa6-7923-9694-e5f8a3005eb0

What is the expected behavior?

The usage doesn't instantly go to 0% ?

Additional information

No response

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

Sudden drop of weekly limit from 70% to 0% #43574

Powered by Codex Action

Same thing happened to me just now!

My Codex usage was showing roughly 80% remaining, and I had checked it only a few seconds before. After one more request, it suddenly dropped straight to 0%.

This happened on 9 September 2026 at around 5:45 PM BST (Europe/London)

It is not related to this:

EDIT: I figured the issue. If you don't renew your subscription, instead of a nice UX optimized message telling you to renew, it simply sets you at 0% without explanation.

I had to use another AI to debug.

Please improve this flow. Thanks!

I went on the $200 plan only a couple of days ago.

Same thing here!
I went from 100% to 0% after 3-4 messages, around 15min work!!!
But I'm on the Plus plan.

Same thing here, they're giving me a heart attack with all this nonsense

It looks like this is a snapshot of how the usage was a couple of days ago, before using a reset. So I think someone accidentally restored part of the database to an earlier point? (Resets are still normal I think, I had one remaining to use and I see that one)

Same here. Instant drop from 70% to 0% :D

What do they expect us to do? Write code with our hands like some kind of peasants

Timeline:

GPT-6 released
strong degradation of GPT-5.6
my 600 credit eaten in 5mins
codex spark weekly usage lasts 45sec (wow)
now.. my usage droped instantly from 70 -> 0

😄

Same here, on the pro plan went from 45% usage to 0% and I'm seeing the same reset as before I used my stored full reset

Many of us with the same problem...

Feedback ID: 01a086fa-14ab-70e3-bdf2-f8ecfff29c94

What is this "feedback ID" you guys post?

Same problem here. 70% to 0% as well.
Feedback ID: no-active-thread-01a08720-fd16-7a11-a0de-0abc6fd8224e

OK usage left came back to normal (100%)

I have the same issue

Same issue for me, dropped to zero from about 50-60% weekly usage.

Yeah same thing here. Just starting work and had ~50% last night. No work overnight. 0% now. Do have a Luna based cronjob that did run a few hours ago without issue, so the drop to 0 must have been since this. Double checked that cronjob Luna session just to make sure it didn't get bumped up to Astra Max or something, but can confirm it was just Luna XHigh as usual. MacOS.

Same incident here on ChatGPT Pro / macOS at approximately the same time.

Codex App: 26.901.51231 (build 8109), bundled codex-cli 0.153.4
Platform: Darwin 25.6.0 arm64 arm
Earlier that day I manually used a stored full usage reset, returning the weekly allowance to 100%
Over the next few hours, normal Codex work consumed only about 10%, leaving approximately 90%
The remaining weekly allowance then abruptly jumped from approximately 90% to 39% within about 30 minutes
At 02:24 KST on 2026-09-10, the usage endpoint reported usedPercent=62 for the 10,080-minute weekly window
Rechecked at 02:29 KST: still usedPercent=62; it has not restored yet
I was mainly using GPT-5.6 Terra with High reasoning effort, but the consumption before the jump was normal and does not explain the sudden 51-percentage-point change

This looks like the post-reset weekly usage ledger may have been restored, overwritten, or recalculated from stale pre-reset state. It is not merely gradual high model consumption.

Feedback/session ID: 01a0872d-6e60-79b1-90ed-bb395576af95

Please correlate the reset event and this session with the account-side usage ledger, and restore any weekly quota lost because of the incorrect post-reset state.

OK usage left came back to normal (100%)

Wow. Happened again! Went from 100% to 0% in 9 mins!

It's restored to my 79% thanks you guys

Same/similar issue on ChatGPT Pro with GPT-6 Astra / Very High reasoning.

Timeline (UTC, Sep 9):

15:40 — approximately 70% weekly allowance remaining
16:32 — approximately 60% remaining
~17:20 — weekly allowance reached 0%

The last ~60 percentage points disappeared within roughly 48 minutes during the same task.

I contacted OpenAI Support and provided the affected thread and timestamps. The case was then escalated to a support specialist.

Important update: approximately 5 minutes after the escalation, my weekly allowance changed from 0% to ~50% remaining.

My 2 stored full usage reset credits were not consumed, so this was not caused by me applying a banked reset.

Current observed sequence:

~70% → ~60% → 0% → ~50% restored

This suggests that at least part of the previous usage depletion may have been incorrect or later reconciled server-side.

Support message:
“Escalated to a support specialist; You can expect a response in the coming days.”

I have previously run comparable long-running tasks, including Ultra reasoning, without similar depletion rates.

It was just fixed, you guys should check also:

It was just fixed, you guys should check also:

Nope, not fixed on my side. Well it has been fixed once, but just after I lost all 100% in 9 minutes, and since then they didn't give the usage back 😡

Adding my affected-user report: I’m a ChatGPT Pro subscriber and remain blocked from using Codex as of 10 September 2026.
My concern is suspected incorrect allowance depletion—not simply a request for higher limits, and not solely a banked reset failing to apply.
The practical impact is significant: my last Codex session left my applications incorrectly configured, and the usage restriction now prevents me using Codex to complete the recovery.
The public reports include both failed banked resets and abrupt losses of remaining allowance. Please investigate those symptoms explicitly rather than assuming that resolving reset redemption addresses every affected account. I’m not claiming they necessarily have different underlying causes.
Please clarify what the 9 September remediation covered, how still-affected accounts can have their usage/reset history checked, and whether allowance lost through verified errors will be restored. Please also provide a time for the next public update.
Other affected users: please react and add your own evidence, including date/time/timezone, plan, client/model, allowance before and after, displayed reset time, and whether/when you used a banked reset. Please keep account credentials and private project content out of public comments.
Related cross-report tracker: #41220.

Since the 9 september depletion bug, I've been afraid of using Astra and came back to Sol. No issue since.
But I would like to know if it's now sure to use Astra.

Same symptom here on ChatGPT Plus, fresh occurrence on September 12, 2026 at about 21:00 CEST.

My Codex 5-hour meter was at 100% remaining. I only resumed a few very small tasks/follow-ups, then shortly afterward the meter was suddenly at 0% remaining / fully exhausted.

I was not using GPT-6 Astra. This was an abrupt drop that does not match the visible workload. Please treat this as another current reproduction of the sudden-to-zero usage issue and correlate the server-side usage ledger for the affected window.

Before assuming the quota was actually consumed, there are now three different confirmed causes producing this exact "instant 73%→0%" symptom. Worth checking in this order:

Lapsed renewal. In Sudden drop of weekly limit from 70% to 0% #43574 a user found that an expired subscription just shows 0% — no "please renew" message. Check your plan status page first; it takes 30 seconds and it's the only cause that's actually good news.
The Sept 9–10 weekly-window oscillation. Astra XHigh from 60%+ to 12% to remaining tokens in less than 5 minutes. #44205 / [Windows][Pro 20x] Weekly usage display jumped 60%→97% while reset timestamp changed #44210 / Weekly quota temporarily jumped from ~3% to 52%, reset date moved Sep 15 → Sep 13, then rolled back and interrupted an active task #44262 document server-side counter swaps with spontaneous recovery hours later. Run /status twice a few minutes apart — if the number moves up on its own, it's the counter, not your consumption. Don't redeem a banked reset while it's moving; it gets eaten by the same broken window.
Real consumption. Only if the number is stable across checks: Astra bills ~2.5× Sol per token, Fast mode another ~2–2.5× on top, so an Astra-XHigh-Fast session can legitimately do what you're seeing.

If you're on a $200 plan and need guaranteed uptime regardless of which counter is misbehaving: the local CLI officially supports [model_providers] in ~/.codex/config.toml — any OpenAI-compatible pay-per-token provider, completely outside the subscription counters:

model = "gpt-5.5"
model_provider = "mizumi"

[model_providers.mizumi]
name = "Mizumi"
base_url = "https://api.mizumi.co/v1"
env_key = "MIZUMI_API_KEY"
wire_api = "responses"

export MIZUMI_API_KEY=..., run codex. wire_api = "responses" is required on CLI ≥ 0.122 (old "chat" tutorials are broken) and the provider must support the Responses API. Delete the top two lines to go back to subscription billing. I run mine through Mizumi; any Responses-compatible provider works identically.

Navigation Menu

Repository navigation

Usage went from ~73% to 0% instantly on the $200/mo plan #44199

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

What subscription do you have?

What platform is your computer?

What issue are you seeing?

What steps can reproduce the bug?

What is the expected behavior?

Additional information

2 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Usage went from ~73% to 0% instantly on the $200/mo plan #44199

Description

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

What subscription do you have?

What platform is your computer?

What issue are you seeing?

What steps can reproduce the bug?

What is the expected behavior?

Additional information

Activity

github-actions commented on Sep 9, 2026

Zakariya1057 commented on Sep 9, 2026

cdenn commented on Sep 9, 2026

djoole commented on Sep 9, 2026

Zakariya1057 commented on Sep 9, 2026

cdenn commented on Sep 9, 2026

montella1507 commented on Sep 9, 2026

Zakariya1057 commented on Sep 9, 2026

montella1507 commented on Sep 9, 2026

navyseal4000 commented on Sep 9, 2026

OnekO commented on Sep 9, 2026

djoole commented on Sep 9, 2026

2 remaining items

aaronldtan commented on Sep 9, 2026

djoole commented on Sep 9, 2026

GGBondBlueWhale commented on Sep 9, 2026

MKsys1337 commented on Sep 9, 2026

vzeazy commented on Sep 9, 2026

xyyelaine commented on Sep 9, 2026

djoole commented on Sep 9, 2026

Zakariya1057 commented on Sep 9, 2026

Sergan2B commented on Sep 9, 2026

cdenn commented on Sep 9, 2026

djoole commented on Sep 9, 2026

matthudson commented on Sep 10, 2026

djoole commented on Sep 11, 2026

taobaibais commented on Sep 12, 2026

mizumico commented on Sep 18, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions