Regarding the closed thread here: Complimentary tokens: GPT-5.4-mini is charged for 1m token group rather than 10m token group
Support is STILL giving me the run around about this. Has anyone been refunded yet? Has support been updated yet that staff has taken ownership of this? What fire do i need to light under whose chair? @ted-at-openai
Happy easter…
I have the same issue. Support is still not acknowledging the problem, even though I’ve pointed to the previous thread (which was closed prematurely). And my request for fixing the billing has been denied without reason.
I’ve pushed back, and now they keep requesting screenshots from my OpenAI dashboard. That’s like calling your bank to ask about a fee, and instead of simply opening up your account in their system and investigating, the agent asks you to provide screenshots of your statement.
This fix should be pretty easy, as they only need to query their logs for particular usage within a certain time window.
Thank you very much for the offer to help.
My support case ID is 07225077.
Just to clarify, I’m not asking for a refund, but only for the usage to be reassigned to the correct service tier (data sharing incentivized tokens). It’s just a small amount of money, but I feel it’s unfair if this isn’t rectified. Ideally, you should apply it to all users who were improperly billed on the days it happened.
Thanks again!
Case Number: 06954985
Okay, will bring it up personally with the team.
@Gille: I will also relist the post for other users to join the conversation but will make a little edit, since I can see why it has been flagged. Hope that’s okay!
ah, yeah, i just edited, but please, yes, edit away if you want. watches the edit* reaction: oh. lol. wow. okay. cool.
exact same email again this morning from support. claiming it’s not eligible. no change.
I went back and forth with them yesterday, but no replies since then.
I do have good news from a reddit thread:
This is from the same employee who posted in the original community thread. At least he seems to have an inside track.
My URL got messed up by the system here.
This is the link to the post:
https://www.reddit.com/r/OpenAI/comments/1s79i9m/api_mistakenly_being_charged_actual_money_when_i/oerhfna/
Since the other forum topic was closed, preventing discussion of “where are the refunds?”, here’s the preposterous communication I got back after submitting a fully fleshed-out report.
To better assist you, could you please provide the following details?
- Org IDs (org-*) / Workspace IDs for affected org(s). - Example API request_ids / router/request IDs and UTC timestamps for affected calls (prefer 2–3 sample requests across different orgs/dates). - Usage Dashboard screenshots or CSV exports showing mini model calls charged to the main pool on the same day. - Billing account IDs and invoice lines showing incorrect charges (if available). - Confirmation of compensated data sharing enrollment and enrollment dates for affected org(s). - Sample cURL/SDK request bodies used (so engineers can reproduce routing and logs). - Confirmation whether other mini/nano models (e.g., gpt-5.2-mini, gpt-5.2-nano) are also impacted.Once we receive your response, we’ll investigate further.
This can be regarded as maximum effort to throw up roadblocks in order to do absolutely nothing from a complete issue report being submitted. This is the output if you were to prompt an AI, “demand information to gather, that would be impossible for a single party to be able to supply, so that we will do nothing and make this report go away”.
OpenAI action required:
Hey everyone, Apologies for this experience. We are looking into these requests and I assure you that refunds for all the impacted users who are eligible will be sent out soon. We are actively working on this, please make sure to include your org-id into the support request so we can review your usage and accordingly provide a refund.
Appreciate the update, but could you define what “soon” means here in actual terms?
Right now it reads a bit like Soon™, which is memorable but not actionable. Can you give a concrete timeframe in days or weeks for refunds, and confirm whether support has now been updated so affected users stop getting contradictory responses?
Kinda seems like an interesting way to take an interest free loan out against paying customers at this point. My spend was exactly $2.58, so I really don’t understand the length of time it’s taking to go through this. Seems like a script you could make to figure out if models were used past a limit in a given timeframe in the tier the user is in and issue refunds in bulk. If you’re relying on user reports to refund, then this is still seeming unethical at the very least to me.
Thanks for the update.
Unfortunately, OpenAI support reps are still denying that any billing errors exist.
They look at the screenshots from my account as well as links to this community thread and the previous one, and say nothing is wrong and that I’m not owed any credit.
Can you please get your support organization on board with what’s happening? I can understand if the billing credit is delayed, but when support denies simple facts over and over, it gets very frustrating for those of us who were affected by an API billing error that wasn’t our fault.
Thank you for any help.
Hey, just a heads-up for anyone who was charged for 5.4-mini: check the Billing Overview for the org where the charges happened.
I think OpenAI may have started issuing the credit. You should also be able to see it on the Credit Grants page. I just checked mine and, unless OpenAI randomly decided they love me and wanted to gift me credits 
, it looks like I got it.
If it hasn’t shown up for you yet, I’d guess it should arrive soon.
Not strictly necessary, but definitely appreciated. The time lost is whatever, honestly. Thanks!
Thanks for the tip! I just checked my billing area and also see a credit. The Credit Grants page shows it dated 1 April.
I was surprised to see it there, as I assumed the billing classification would be corrected under the Usage page. But maybe issuing credits was easier for OpenAI than changing the way tokens were already classified.
hmm, randomly decided to check on this thread today… I quietly got credits as well. No correspondence at all about it though… just… sigh…
