Codex credits are draining while idle / not actively used

Codex
1

Sorry for starting a new topic if this is already being tracked somewhere. I searched and found some related threads, including “Codex Credits Dissappearing”, “Codex usage after the limit reset update”, and reports where a small number of Codex tasks consumed a large part of the limit/credits.

However, I am trying to find whether there is an active topic specifically about Codex credits draining while Codex is idle or not actively being used.

In my case, the main problem is not simply that a large task used many credits. The problem is that credits appear to keep disappearing while Codex is left open/running, without me actively starting new meaningful work.

This happens repeatedly. Every night, when Codex is left running, credits continue to drain. It is still happening. Every time I step away from the computer and later come back, I find that more credits have disappeared.

My workflow has not materially changed compared with before. I use Codex in a similar way as before, including as a supervisor/monitor for my home system. But since GPT-5.5 became available, both the weekly included credits/limits and purchased credits are being consumed at an extreme and unpredictable rate.

The weekly credits/limits started disappearing so quickly that I was genuinely shocked. I then started paying for extra credits simply so I could continue using Codex. But the purchased credits also began disappearing in the same abnormal way.

Recent OpenAI payments from my bank statement:

  • April 23, 2026: OPENAI *CHATGPT SUBSCR, DUBLIN - 23.00 EUR
  • April 28, 2026: OPENAI* CHATGPT CREDIT, DUBLIN - 40.00 EUR
  • May 2, 2026: OPENAI DUBLIN Co.IE - 5.72 EUR
  • May 3, 2026: OPENAI DUBLIN Co.IE - 10.04 EUR

Recent credit-drain example:

  • Today I bought 250 credits.
  • The balance dropped from 250 to 130.
  • Later it dropped from 130 to 64.
  • I did not run anything unusually large or special that would explain this.
  • I also deliberately used Codex to draft this post so I could observe the credit consumption in real time.
  • I started this drafting session with 64 credits.
  • By the time I reached this point, I was down to 18 credits.
  • That means 46 credits were consumed just while drafting/reporting this issue.

There was also an overnight case where Codex was left running in the same monitoring/supervision role as before, and around 20% of credits were consumed by the morning. This is not an isolated case. The same pattern happens every night and every time I leave Codex open while I am away from the computer.

The most important issue is idle/background credit leakage. If Codex is doing background model calls, retries, summaries, monitoring actions, tool calls, context refreshes, or anything else billable while the user is not actively using it, that needs to be clearly visible and controllable.

Right now Codex does not provide enough detail for me to verify what consumed the credits. I cannot see a clear per-task or per-session breakdown explaining why credits were consumed while Codex was just left open/running.

Has anyone else seen this specific issue: Codex credits draining while idle or not actively used?

Also, where is the correct place to file a formal billing complaint or request credit restoration/refund for this? The in-app Codex Feedback form looks like a bug report form, not a real billing support channel. I need someone from OpenAI to inspect internal logs and explain what triggered the charges.

Please, help.

2

Welcome to the forum!

I also couldn’t find an active similar topic that seemed to cover this exact “credits draining while Codex is idle/ not running” issue.

For billing/credit restoration, I’d recommend contacting the Help Center directly:

https://help.openai.com/en

You may also want to check the Codex rate card, in case it helps with understanding how credits are counted:

https://help.openai.com/en/articles/20001106-codex-rate-card

That said, if credits are being consumed while you are not actively starting tasks, that sounds like something Support would need to investigate from the account/session side.

3 
Thank you for the reply.

I followed your suggestion and contacted OpenAI Support by email with a detailed explanation of the issue

I asked Support to investigate this from internal account/session logs, because Codex does not expose enough per-session billing detail for users to verify what triggered the charges.

I’ll update this topic if I receive a useful response, in case it helps anyone else seeing similar idle/background credit consumption.

Thanks again...
4

Good!

And yes, please keep us updated here.

Hopefully you get answers soon and that it will be fixed!

5

Update:

I contacted OpenAI Support about this issue.

The response I received did not provide a billing/session-level explanation of what consumed the credits. It also did not address the main practical problem I reported: that Codex credit usage suddenly became dramatically higher than before for the same workflow.

Support suggested that I record my screen while the issue is happening, in order to show credits being consumed by themselves. However, this does not really address the core issue.

The problem is not only that credits appear to be consumed while Codex is idle or left open in the background. The problem is also that, after GPT-5.5 became available, my Codex usage seems to consume credits roughly 5x faster than it did before, even though my usage pattern has not materially changed.

In other words, the support response focused on proving/recording the “credits draining by themselves” part, but did not address the much broader issue: the abnormal increase in consumption rate compared with previous Codex usage.

What I still need from OpenAI is a concrete explanation of:

  • which Codex sessions/tasks consumed the credits
  • whether any background activity, retry loop, context refresh, summarization, tool call, or monitoring process consumed credits
  • whether GPT-5.5 or another more expensive model was automatically selected
  • why the same workflow now consumes credits at a much higher rate than before
  • whether this is expected behavior, a bug, or an unintended billing/product-side change
  • whether affected users can receive a credit restoration/refund

At the moment, I still do not have a clear explanation for the sudden 5x increase in Codex credit consumption or for the apparent idle/background drain.

6

Thanks for taking the time to update us, even though you haven’t yet got clear answers why it’s happening.

Support might even reply to you here in your topic.

7

I’ve got 2 bots running 16 hours per day everyday. Both running 5.3 codex in CLI. 10000 credits lasts me about 5 days. I just ran out and recharged again and it’s on day 5 so it’s still about the same for me. I think what you need to do is constantly start new sessions like I do every 1 million tokens spent within a session otherwise you start to incur what I’ve started to call token taxes. You just need to keep all you work planned and tracked in local docs instead of relying on the models fragile context window. If you’re still burning tokens when at idle then that means something else is still running.

8

When you say “credits are being used” it shoudl be noted that the dashboard is not instant, there is some delay to it, so you might be seeing events that occured prior, the real test is to stop doiung something then wait a couple of hours and see if you are still using tokens.

Also check to see if you have OpenClaw running as that can have pings that consume tokens in the background.

Let em know if you still have issues.

9

To be clear I have been having this issue also. And I can even be 100% sure:

this happened to me directly after a fresh reset, my usage was at 100% hourly and weekly, I was NOT using codex, but I had it open because I was planning to. I was reading the current diff to remember what I was working on previously, and then when I got to codex, my hourly and weekly usage drained – hourly down to 80%.

I, still having not used codex in a few days(no messages at all), deciding to start writing a post. ( What automatically consumes Codex usage without doing anything? - #5 by EricGT )

While I was writing that post, I observed the usage further going down, despite no usage being done yet.

10

I think if you have memories enabled, that may be what it is. I had the issue, turning it off SEEMED to fix it.

11

Here is a general update for everyone following this.

I received a reply from OpenAI Support. They said they will proceed with a deeper internal review of my account and usage, specifically the session where my Codex credits dropped from 64 to 1 during what was, from my side, normal drafting activity.

They also said they will check whether a higher-cost model, such as GPT-5.5, was automatically used, and whether background processes such as retries, context handling, or tool-related calls may have contributed to the observed credit usage.

They explicitly acknowledged that this level of visibility is not currently exposed on the user side, and that investigating it requires internal review.

Their temporary recommendations were:

  • monitor which model is selected, if visible

  • start a fresh session to isolate behavior

  • avoid prolonged sessions with large context, since that can increase usage

A few clarifications from my side:

@Foxalabs I am not using OpenClaw or any third-party client/wrapper. This is about the official Codex app.

I am fully aware that the dashboard is not instant and that usage reporting can lag. That is not the issue. Even taking delayed reporting into account, the behavior is not consistent with how Codex behaved for me until now.

My workflow has not meaningfully changed. Same kind of tasks, same kind of sessions, same way of using Codex. What changed is that ordinary procedures that previously used very few credits now appear to consume credits much faster and in a much less predictable way.

Also, even when I manually leave Codex set to GPT-5.4, which I had been using successfully and was satisfied with, the app often defaults back to GPT-5.5. I would estimate this happens in most cases. That makes usage harder to control because I have to keep checking the selected model every time.

Regarding the Memories point raised by @Noah13: thank you, that was useful. I also passed this information to OpenAI Support.

In my case, Memories were enabled on at least two machines without me intentionally enabling them. On a third machine they were not enabled. Since turning Memories off, credits no longer seem to be draining “by themselves”, or at least not anywhere near the same level as when Memories were enabled.

Of course, I cannot prove causality from my side. But this gave me a concrete direction to test, and the behavior after disabling Memories does seem different.

@Noah13’s latest description is also very relevant: after a fresh reset, usage dropped significantly even though Codex was apparently not being actively used, only left open while reading a diff. That is exactly the kind of behavior that makes this difficult to understand from the user side. It suggests the issue may not be limited only to visible prompts or explicit user actions.

This is also why I keep saying this is not just a subjective feeling. Subjectively, yes, it feels like credits are being consumed much faster and less predictably than before. Objectively, my credit balance and actual spending confirm that something has changed.

The problem is not that Codex costs credits. That is understood. The problem is lack of visibility and predictability. If credits are being consumed by model switching, large context, retries, Memories, background processes, tool-related calls, or idle/open sessions, users need to be able to see that clearly.

At this point, I have almost developed a conditioned reflex of anxiety when opening Codex, because I no longer know whether a simple task, a long context, an idle window, Memories, automatic model switching, retries, or some background process will start consuming credits in a way I cannot inspect or control.

I hope OpenAI’s review clarifies what happened and that this leads to better transparency for all users, not only for my specific case.

Update from today:

This morning I paid for another 250 credits.

In less than an hour, I went down to around 150 credits while doing the same type of work and the same processes I had been doing before.

Then I started one more task with 150 credits remaining. That task was not a new complex project or anything unusual — it was basically maintaining/adjusting one existing procedure.

That single operation took me from 150 credits down to 71 credits.

This was while using GPT-5.4, not GPT-5.5.

So again, this is not about me suddenly changing my workflow or doing something dramatically different. These are the same kinds of Codex tasks I was doing before, but the credit consumption is now massively higher and much harder to predict.

12

Hey everyone, Credits should map to Codex usage/model activity, not when codex is in idle state. The current rate card is token-based, so larger context, retries, background tasks, model choice, fast mode, or automations can make usage jump even when there are fewer visible prompts.

Since this involves account/billing details, Support will need to review the account/session records. @Zormic, since you already have a Support request open, can you share just the case ID or case subject here so we can track the issue and work on resolving it for you?

To help isolate it, can you share in the request the exact time window where the balance dropped, with before/after screenshots from Codex Settings → Usage, and whether Memories, background terminals/subagents, cloud tasks, CLI/IDE sessions, or automations were active during that window?

13

I wanted to tap in and also share my similar experience.

I bought 1000 credits yesterday after my quota was depleted, but I haven’t used any of them atall. After around an hour, they were ALL gone on their own. I contacted the support, and after a couple of emails back and forth, they restored my credits balance. But today my credits started being consumed on their own again, and now my balance is showing 993. (For reference, I do not have any active sessions, haven’t sent a single message, no automations, basically zero codex activity)

Hope this gets real attention and resolves quickly.

edit 16:16 UTC: 931 credits left (still haven’t used codex at all, they are draining on their own)

14

Hi OpenAI Support,

Sorry for the delayed reply. I am currently traveling, so I cannot always respond promptly, and my internet connection is not always stable.

My case number is: 08420955

Regarding the additional details you asked for: I understand why you need exact time windows, before/after screenshots, and information about Memories, background terminals/subagents, cloud tasks, CLI/IDE sessions, or automations. I will try to provide as much of that as possible when I can.

However, being on the road makes this kind of “detective work” more difficult at the moment. Also, respectfully, OpenAI should have significantly better visibility into what actually happened on the account/session side than I do as an end user.

From my side, the issue is both subjective and objective.

Subjectively, Codex feels much less predictable and much more expensive to use than before.

Objectively, my credit balance and my actual spending show a very clear change. Since GPT-5.5 appeared, my impression is that credit consumption has increased drastically. My wallet confirms that this is not just a feeling.

For example, yesterday morning I paid for another 250 credits. In less than an hour, I dropped to around 150 credits while doing the same kind of work and the same processes I had been doing before. Then I started one more task with 150 credits remaining — not a new complex project, but basically maintaining/adjusting an existing procedure — and that alone took me from 150 credits down to 71 credits. This was while using GPT-5.4.

Again, I understand that larger context, retries, background tasks, model choice, fast mode, automations, etc. can affect usage. But the point is that this behavior is not consistent with how Codex behaved for me until recently, while my workflow has not meaningfully changed.

I hope the open support case gives you enough internal visibility to track the relevant sessions and understand what changed.

15

Same here.

At one point, 1000 credits melted away in just a few hours. I was honestly shocked by how fast they disappeared.

After that, I kept buying smaller credit packages just so I could continue working and keep at least some balance available, but even those credits were draining abnormally fast.

So your case sounds very familiar to me. The difference is that in my case I was using Codex during some of those periods, but the rate of consumption was completely out of proportion compared to the same workflow I had before.

16

I too also had this issue. I only purchased credits for the first time to test it out. 1000 Credits were present yesterday. This morning I get up and check and its 0 despite almost zero usage.

Under Credit Usage history is also blank

image
image2029×614 20.7 KB

I hope this matter gets resolved quickly

17

Full forum update

I also have a small widget I made to monitor Codex usage/credits, because the behavior became impossible to follow manually.

The screenshots show the same thing from two sides:

  • the Codex Usage page itself

  • my own monitoring widget, which tracks the same values over time

At the time shown, I still had:

  • Weekly usage available

  • 5 hour usage limit available

  • paid credits remaining

Then, suddenly, the paid credits started disappearing, even though nobody was actively doing anything.

This happened overnight. No active work. No deliberate Codex usage. Nothing that should explain this kind of paid-credit drain.

One more clarification: I logged in remotely to my machine, as I mentioned earlier, only to check what was happening.

I am currently traveling, so I am not actively working in Codex. I was definitely not doing any Codex work during that period. I only connected remotely afterward to verify the state and capture/check the usage information.

The most worrying part is that this is not happening when the normal usage limits are exhausted. The screenshots show that both the weekly limit and the 5-hour limit still had availability, yet the paid credits were still being consumed.

That makes the situation even harder to understand from the user side.

From my perspective, the credits are simply leaking everywhere. I am not talking about a delayed dashboard update or normal token-based usage. I am talking about paid credits disappearing while there is still available quota and no active work being done.

This is exactly why I keep saying the issue is not just “Codex costs credits.” The issue is that the consumption is no longer predictable, visible, or controllable.

SCR-20260506-hfza-2
SCR-20260506-hfza-22876×1900 242 KB

18

The CODEX app consumed 1500 credits from my account while I was sleeping. When I sat down at my PC, I saw it charging 300 credits in increments of 100, even though I hadn’t used anything. I disabled automatic billing and checked, and it had stopped charging in increments of 100 and started consuming credits in increments of 10, even though I hadn’t used any prompts. I was researching this, and it seems several other people had the same problem. When my account ran out of credits, it charged my credit card until I disabled automatic billing.

19

Hey @gideon100 , Can you please open a support case with us and provide some details. Exact time, before and after credit usage and specially details relating to your account so we can review the usage. Please do share the case id/case subject here with us, so we can track the request. Thank you!

20

same here, 25 terminals running!?