Hi,
I’m posting this because I have been stuck in a repetitive support loop and still cannot get a basic accounting reconciliation for 2,500 Codex/Work credits that I purchased.
On September 20, 2026, I reached my included Codex/Work usage limit.
I purchased 2,500 additional credits so that I could continue my development work.
The purchase was successfully completed. My invoice clearly shows:
Credits — Quantity 2,500
I then resumed my Codex/Work task.
A short time later, Codex stopped again and told me that I had no credits remaining and needed to purchase more credits.
I contacted OpenAI Support because consuming 2,500 credits in such a short period seemed unusual, and I wanted to see the actual credit usage records.
Support eventually provided what it described as the Codex credit consumption ledger entries during the one-hour period after my payment.
According to Support, my payment occurred at approximately:
2026-09-20 10:20 UTC
They then provided the following consumption entries between approximately 11:00 and 11:05 UTC:
-0.064443
-25.143625
-9.296875
-0.142471
-7.355375
-7.553375
-11.504625
-11.447000
-0.088709
-0.124862
-9.979125
-0.055891
-9.832750
-9.574500
-12.219875
-0.162726
-9.419250
-0.118090
-11.416375
-0.077593
-0.078908
-11.544375
-10.840875
-3.302331
-7.807044
Support also stated that -7.807044 is my current negative balance.
Here is the accounting problem:
The total of all of the consumption entries provided by Support is only approximately 169.151 credits.
I purchased:
+2,500 credits
My current balance is supposedly:
-7.807 credits
Therefore, approximately 2,507.807 credits of total consumption or adjustment would need to be accounted for in order to reconcile the balance.
But the ledger entries that Support itself provided add up to only approximately:
169.151 credits
That leaves approximately:
2,338.656 credits completely unexplained.
This is the specific issue I am asking OpenAI to investigate.
I understand that Codex usage can vary significantly depending on the model, reasoning effort, task complexity, context size, tools, sub-agents, and background activity.
But that does not resolve the accounting discrepancy.
If another 2,338+ credits were legitimately consumed by sub-agents, background tasks, tool activity, or some other billable process, then those charges should exist somewhere in the account-level billing or credit ledger.
I am not disputing that my task used credits.
I am asking where the remaining ~2,339 credits are recorded.
OpenAI Support keeps asking me questions such as which model I used, which mode I used, whether there was one task or multiple tasks, etc.
I have tried to cooperate with these questions, but the conversation keeps returning to the same generic explanations without answering the numerical discrepancy above.
At this point, the primary issue is not the model or mode.
The primary issue is very simple:
2,500 purchased credits + a final balance of -7.807 requires approximately 2,507.807 credits of accounted usage.
The usage ledger provided to me accounts for only approximately 169.151 credits.
Where are the ledger entries for the remaining approximately 2,338.656 credits?
If there was a pre-existing negative balance, please show the exact balance before my purchase and the ledger entries that created it.
If there were additional background tasks, sub-agents, tool executions, or other billable activity, please show those transactions and their timestamps.
If the displayed ledger is incomplete, I need someone with access to the actual backend billing/credit ledger to review the account and reconcile the full 2,500-credit purchase.
I am not asking for another general explanation of how Codex pricing works.
I am asking for a straightforward accounting reconciliation:
Starting balance after purchase: +2,500 credits
All debits/adjustments: ?
Final balance: -7.807 credits
The debits shown to me so far do not mathematically reconcile those numbers.
This issue is currently blocking my development work, and I do not feel comfortable purchasing another 2,500 credits while more than 2,300 credits from the previous purchase remain unexplained.
I would greatly appreciate escalation to an OpenAI employee or team that can access the actual Codex/Work billing ledger and investigate this urgently.
I have screenshots of the invoice, credit balance, usage dashboard, and the Support conversation containing the ledger entries above.
Thank you.