Account-specific model access failures across Codex clients; no route to human technical support

Hello OpenAI team/community,

I’m still experiencing persistent, account-specific failures when trying to use models. The issue reproduces with any model, and the error is not consistent, including capacity-style messages such as “overloaded.” It occurs on the same device, network, and conversation thread in both Pi and the official Codex client.

For comparison, a friend’s Pro 5x account works under the same conditions, while my Pro 20x account fails. We also swapped accounts between clients, and the failures continued to follow my account rather than the device, network, client, or a specific model.

The issue has also affected ChatGPT Web conversations and Dot tasks. Tasks may stop midway, the thinking state may disappear, and the visible active state may remain even when no further action is happening.

I have been unable to find a path to human technical support. The official support channels route me to automated responses, I cannot create a technical support case, and I cannot reach a human agent. Could you investigate whether an account-level entitlement, routing, quota, or backend flag is causing this, and advise how I can obtain an account review?

I can provide timestamps, request IDs, client/version details, and sanitized logs privately if needed. Thank you.

Privacy: I will not post email addresses, account or organization IDs, API keys, tokens, private prompts, conversation content, IP/device fingerprints, or unredacted screenshots publicly.

Thanks for sharing such a detailed report. Your troubleshooting is particularly useful because you’ve isolated the issue across multiple clients, models, and environments.

The fact that failures consistently follow the account strongly suggests that further device- or network-level troubleshooting may not be productive. This deserves an account-level technical investigation.

I checked OpenAI’s official documentation and found two potentially useful references:

  1. Troubleshooting model access — explains possible account restrictions, workspace permissions, and usage-related limitations.

  2. Official support incident — confirms that support@openai.com has been provided as an alternative contact channel when Help Center chat support is unavailable.

Since you’ve already tried the standard support channels, emailing a concise technical report directly may be worth attempting. Include your timestamps, sanitized request IDs, exact errors, and the results of your account-level comparison, and explicitly request an investigation and a trackable support reference.

More broadly, I think your report highlights an important product issue: users experiencing persistent, reproducible account-specific failures need a reliable escalation path beyond automated troubleshooting.

I hope someone from the appropriate team can review your case.

Hey @spongebob_261001, thanks for documenting the checks, and @techsoft.ye for adding your input.

Are these failures still happening? If so, please add a recent request ID, the exact error, its timestamp and time zone, and your Codex client version to your existing support request. If you don’t have one, please email those details to support@openai.com .

Once submitted, share just the case number here so I can review the details.

- Moizz