Codex unavailable across Web, CLI and ChatGPT Work - "Selected model is at capacity" on every model

What version of the Codex App are you using (From “About Codex” dialog)?

N/A - Issue also reproduces in Codex Web, ChatGPT Work and Codex CLI 0.149.1

What subscription do you have?

ChatGPT Plus

What platform is your computer?

Get-ComputerInfo | Select-Object WindowsProductName, WindowsVersion, OsArchitecture

What issue are you seeing?

Codex is currently completely unusable on my account.

Every attempt fails with the following error:

Selected model is at capacity. Please try a different model.

The important point is that this is NOT limited to one Codex client or one model.

I have reproduced the exact same problem in:

• Codex Web
• Codex CLI
• ChatGPT Work / Codex tasks

I have also tested multiple available models and they all fail with the same error.

Codex CLI information

Codex CLI version:

codex-cli 0.149.1

The CLI starts correctly and recognizes my ChatGPT Plus account.

My /status currently shows approximately:

• 5-hour limit: 85% left
• Weekly limit: 54% left

Therefore, my Codex quota is not exhausted.

Reproduction

1. Open Codex Web, Codex CLI, or ChatGPT Work.
2. Start a new session/task.
3. Select any available model.
4. Send even a minimal prompt such as Hello.
5. The request immediately fails with:

Selected model is at capacity. Please try a different model.

Changing the model does not resolve the issue.

Expected behavior

Codex should process the request normally because my account still has substantial usage quota available.

Actual behavior

Every Codex access method associated with my account fails, regardless of the selected model.

Because the same problem occurs in Codex Web, CLI, and ChatGPT Work, this does not appear to be a local installation, project, Windows, Node.js, or CLI-specific issue.

It may be related to account-level routing, capacity allocation, admission control, or a server_overloaded state affecting the account.

Please investigate whether there is an account-level backend/routing issue.

This issue currently prevents me from using Codex entirely.

What steps can reproduce the bug?

1. Sign in to my ChatGPT Plus account.

2. Open Codex using any of the available access methods:

   • Codex Web
   • Codex CLI
   • ChatGPT Work / Codex tasks

3. Start a new session/task.

4. Select any available model.

5. Send a minimal prompt such as:

   Hello

6. The request immediately fails with:

   Selected model is at capacity. Please try a different model.

7. Change to another available model and retry.

8. The exact same error occurs again.

The issue reproduces consistently across Codex Web, Codex CLI, and ChatGPT Work.

Codex CLI version:
codex-cli 0.149.1

Current usage shown by /status:

• 5-hour limit: 85% left
• Weekly limit: 54% left

Therefore, the issue does not appear to be caused by exhausted usage limits.

The issue also reproduces with completely new sessions and very small prompts, so it is not related to context-window usage or a specific project.

What is the expected behavior?

Codex should accept the request and generate a response using an available model.

If a specific model is temporarily at capacity, selecting another available model should allow the request to continue.

Since my account still has substantial Codex usage available, Codex should not be completely unavailable across every model and every access method.

Additional information

This issue appears to affect my account across the Codex backend rather than a specific client.

The same behavior occurs in:

• Codex Web
• Codex CLI
• ChatGPT Work / Codex tasks

It also occurs with every model I have tested.

Because the same issue reproduces outside the CLI, it does not appear to be caused by my local Windows installation, Node.js environment, CLI version, repository, or project configuration.

My account still has available usage:

• 85% of the 5-hour limit remaining
• 54% of the weekly limit remaining

Please investigate whether my account is affected by an account-level routing, admission-control, capacity allocation, or server_overloaded condition.

At the moment, Codex is completely unusable on my account across all available access methods.

GitHub suggested the following potential duplicates:

• #45901 - Persistent "Selected model is at capacity" error across multiple models in Codex CLI, while other accounts work
• #44531 - Intermittent "Selected model is at capacity" errors in Codex Desktop
• #41790 - Repeated "Selected model is at capacity" errors interrupting Codex tasks

My case appears broader than these reports because the same account consistently fails across:

• Codex Web
• Codex CLI
• ChatGPT Work / Codex tasks
• Multiple available models

The error is not intermittent and is not limited to one Codex client.

This may therefore be an account-level backend/routing issue rather than a client-specific capacity problem.

No repository file or test is named. Start by reproducing the failure through Codex Web, Codex CLI 0.149.1, and ChatGPT Work, checking the CLI /status output and comparing related issues #45901, #44531, and #41790. Done means requests succeed across the affected access methods and models for the reported account.