Codex model not supported: account check

What you see

Codex lets you choose or configure a model, but a request fails with a response saying that model is not supported with a ChatGPT account. Save the exact model name and full error. A historical report shows this wording for gpt-5.4; it does not prove the same model is the one failing for you now.

This message concerns model access. If the response instead says “Selected model is at capacity”, that is a different symptom. A remaining-usage percentage does not by itself establish access to every model.

Check the access boundary

1. Note whether you are using the Codex CLI, desktop app or IDE extension. Confirm which ChatGPT account or workspace the affected client displays. An API-key session and a ChatGPT-sign-in session use different access arrangements.
2. Look up the named model in the current Codex changelog. A rollout announcement may limit availability by plan, client or workspace. In particular, gpt-5.4 and gpt-5.4-mini retired from ChatGPT-signed-in Codex on August 31, 2026; the historical gpt-5.4 error above predates that retirement. API-key access has separate rules.
3. If you are in an Enterprise workspace, ask an administrator to check its model availability settings. GPT-6 Sol and Luna were off by default for Enterprise at launch, and selecting a model locally cannot enable one your workspace has not allowed.
4. In the CLI, the documented /model picker can show your current choices. /status confirms what the session selected. A model appearing in a picker or configuration is not proof that the next request will be accepted for this account.

If the rejected model is pinned in a saved configuration or task, check that setting against the currently available models. Record the original value before editing; change only the model selection that you have identified. An update guide can help locate an outdated installation, but updating alone is not a guaranteed account-access repair.

If an apparently available model still fails

Compare one short request in the same account and client using a model the current documentation says is available to that surface. If the error persists across available models, preserve the exact response, time, client version, account plan and workspace name for support. Do not post tokens or authentication files.

One public report described several models rejected on a paid account. The reporter suspected account entitlement synchronization, but that remains a hypothesis. Their API-key comparison does not mean switching to API-key billing repairs ChatGPT subscription access. Deleting local sign-in data or repeatedly logging out also lacks support as a general fix for this error.

How to tell it worked

The selected model completes a small request in the affected client and sign-in method. Seeing a model name in a settings panel or a successful request through a different billing method is a narrower result.

Known limits of this guidance

This guide combines current official access rules with issue reports from particular versions and accounts. It cannot identify the cause of every rejection from the error text alone. The March and September reports are evidence of reported symptoms, not proof of an ongoing incident or a universal workaround.

This guide is part of our Codex sessions troubleshooting.

First check whether this is a Codex service-side problem → View live Codex status.