We’re a small team on a newly purchased ChatGPT Business plan (2 ChatGPT seats), using Codex via the desktop app. We’re hitting a workspace-wide block that doesn’t match the usage we’re seeing, and support guidance has been contradictory, so I’m hoping someone from the team or community can help clarify.
What’s happening
With very light use, our whole workspace gets blocked: “You’ve reached your workspace credit limit / Your workspace is out of credits,” sometimes cutting off a task mid-run.
This has now happened on two consecutive days.
The numbers (whole workspace, both users combined)
Billing: 2/2 ChatGPT seats (“ChatGPT and Codex, $25/mo”), 0 Codex seats, credits balance 0, no monthly usage allocations set.
Why this is confusing
Support told us Business has a fixed daily Codex quota where only successful turns count and auto-review doesn’t — but if so, 13 successful turns exhausting the entire workspace’s daily allowance seems far too low and looks like a misprovisioned quota or a bug.
Our admin billing page shows a banner: “Add credits so members can keep working without interruption” — which contradicts support telling us that adding credits does NOT raise the quota.
The usage dashboard shows 0 credits/0 tokens consumed even while the limit is being enforced.
Questions
What is the actual included Codex allowance per ChatGPT seat on Business (in requests/turns per rolling window)?
Is ~13 turns/day exhausting a Business workspace expected, or a misconfiguration/bug?
Does adding credits actually restore Codex access on Business, or not? (The banner and support guidance conflict.)
Has anyone else on Business seen the dashboard show 0 usage while being blocked as “out of credits”?
Thanks in advance — any definitive answer on the included Codex quota for Business ChatGPT seats would help a lot.
Thanks for sharing the details @Info-Yingzhong. I can see why this is confusing. Also good points from @PaulBellow and @mgm2.
I agree with @PaulBellow that not all Codex turns are equal. A single turn can consume far more resources than another depending on context, files, and tool usage.
That said, a 2-seat Business workspace hitting a block after only ~13 successful turns does seem unusually low. The bigger red flag is the mismatch between:
Workspace blocked as "out of credits"
Usage export showing 0 credits used
Billing page encouraging credit purchases
Support saying credits don't increase the included allowance
That combination feels more like a metering or provisioning issue than normal usage limits.
OpenAI also doesn't publicly describe Business Codex access as a simple "X turns per day" quota, so it's hard to reconcile the behavior you're seeing with the reported usage.
If you're comfortable sharing it, drop the support case ID here. That would make it easier to investigate further and see if there are any irregularities with the workspace configuration or quota enforcement.
I have problems aswell… The usage shown is not true… first the weekly limit should have resettet at another day… then ist was resettet before that day…
but with that reset i was out of tokens within 2 days… even used medium… before i did alot more with codex and played long sessions with Paperclip… nothing within this period…
What happened? I’ve been working on the same project in the exact same way every day for about two weeks… I’d say that, as a result, prices have shot up more than tenfold overnight?
Thanks Mark G. and everyone — that summary matches our read exactly: the mismatch between being blocked as “out of credits,” the export showing 0 used, the billing page pushing credit purchases, and support saying credits don’t raise the allowance points to a metering/provisioning issue rather than normal usage.
Happy to share the case ID for investigation: Case #09534429.
A quick update since the original post: the limit recurred again mid-task even after support told us it had “reset,” and we’ve now received three different explanations on the case (a fixed daily Codex quota where credits don’t help, then the billing page saying credits DO restore access, then most recently that it was a GPT-5.5 Pro model allocation tracked separately from credits). We run comparable tasks on the same model on a personal Plus account without hitting this wall nearly as fast, so a 2-seat Business workspace being more restrictive than a single Plus seat is exactly the part that doesn’t add up.
If anyone with visibility into workspace quota provisioning can check #09534429 for an enforcement/metering irregularity, that would be hugely appreciated. We’d like to stay on Business — we just need the workspace correctly provisioned. Thanks again.
Thanks for the additional details @Info-Yingzhong, and for sharing the case ID. I can see why the changing explanations have made this difficult to pin down.
One thing that may help separate the different signals you're seeing: the workspace credit system and the model usage limits are not the same thing.
The current behavior described in OpenAI's flexible pricing documentation is:
Users receive access to ChatGPT and included model usage through their seat.
Certain advanced model usage and features can be subject to usage limits (for example, rolling-window limits such as a 5-hour allocation).
Credits are separate from those included limits.
You do not need credits to use core ChatGPT functionality such as standard chat, Search, file uploads, or Canvas.
If a user reaches an included usage limit and the workspace has purchased credits, usage can continue by drawing from the workspace's shared credit pool.
Purchasing credit packs is optional unless a workspace wants to continue usage beyond included limits or is using features that consume workspace credits (such as Codex seats).
The confusing part in your case is that the error messaging appears to be presenting a workspace credit-style message ("out of credits") while support has also referenced what sounds more like a model allocation or rolling usage limit. Those are different enforcement mechanisms, which may explain why the dashboard can show 0 credits consumed while access is still being restricted.
Kindly check the usage details here: https://admin.openai.com/analytics/codex
