Codex Desktop: All models report "Selected model is at capacity" on one Pro account, while another account works on the same Mac #52894

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

26.1007.21159

What subscription do you have?

ChatGPT Pro 20x

What platform is your computer?

Darwin 27.0.0 arm64 arm

What issue are you seeing?

Description

Codex Desktop consistently reports the following error across all available models on my personal ChatGPT Pro account:

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

However, another personal ChatGPT account works normally on the same Mac, network, and Codex Desktop installation.

The affected account has its full usage allowance available and Codex Web works normally.

The issue persisted for several days, temporarily resolved for approximately one day, and then returned.

Steps to Reproduce

1. Open Codex Desktop and sign in with the affected Pro account.
2. Start a new conversation and submit a simple prompt, such as Reply with OK.
3. Observe the capacity error across all available models.
4. Switch to another personal ChatGPT account in the same desktop application.
5. Submit the same prompt.
6. The second account works normally.

Expected Behavior

Codex Desktop should process requests normally when the account has available usage allowance.

Actual Behavior

All models fail on the affected account while another account works on the same machine.

Neither account belongs to a managed Business, Enterprise, or Edu workspace.

Relevant Logs

Previous logs showed:

WebSocket was closed before the connection was established
closeCode=1006
SSH: codex path probe timed out after 60000ms
app_server_connection.connection_failed_before_ready
app_server_connection.reconnect_failed

Repeated connection failures were observed, although their relationship to the capacity error has not been confirmed.

Feedback Upload Issue

I attempted to submit in-app feedback including diagnostic logs.

Feedback ID:

01a12321-cbc7-7e12-8fd5-320a3534087a

However, Codex Desktop displayed:

Feedback upload not confirmed. We couldn't confirm whether your feedback was received. It may have been delivered, and resubmitting could create a duplicate.

Therefore, the feedback upload status remains unknown.

Additional Information

• Codex Desktop version: 26.1007.21159
• Subscription: ChatGPT Pro
• Usage: Full allowance available
• Both accounts: Personal accounts
• Codex Web: Working on the affected account
• Failure pattern: Recovered for approximately one day, then recurred
• Switching models: Does not resolve the issue

Please investigate whether this is related to account-specific model access, backend request admission, routing, authentication state, or execution-environment problems.

Thank you.

What steps can reproduce the bug?

Steps to Reproduce

1. Open Codex Desktop version 26.1007.21159 on macOS.

2. Sign in with my affected personal ChatGPT Pro account.

3. Start a new conversation.

4. Select any available model.

5. Enter a simple prompt:

   Reply with OK

6. Submit the request.

7. Codex Desktop returns:

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

8. Repeat with other available models and new conversations. The same error occurs.

9. Sign out and switch to a different personal ChatGPT account on the exact same Mac, network, and Codex Desktop installation.

10. Submit the same prompt. The second account works normally.

11. Switch back to the affected account. The capacity error occurs again.

Reproducibility

• The issue has occurred repeatedly over several days.
• It temporarily resolved for approximately one day before recurring.
• All available models are affected on the original account.
• The affected account works normally in Codex Web.
• Neither account belongs to a managed Business, Enterprise, or Edu workspace.

Session / Feedback Information

• Feedback ID: 01a12321-cbc7-7e12-8fd5-320a3534087a
• Feedback upload status: Unconfirmed by Codex Desktop.
• Usage allowance: Fully available; no usage limit reached.
• Token limit usage: Not available.
• Context window usage: Not available. The error also occurs in newly created conversations with a minimal prompt.

No special code or repository setup is required to reproduce this issue. A simple text prompt is sufficient.

What is the expected behavior?

Codex Desktop should successfully process prompts and generate responses using the available models when the account has sufficient usage allowance.

Both personal ChatGPT accounts should be able to use Codex Desktop normally under the same device and network conditions.

For example, submitting Reply with OK in a new conversation should return a response rather than a model capacity error.

The affected account should behave consistently with its working Codex Web access and the other functioning account.

Additional information

Environment

• Codex Desktop version: 26.1007.21159
• Operating system: macOS
• Subscription: ChatGPT Pro
• Workspace: Personal account (no managed workspace)
• Usage allowance: Fully available
• Codex Web: Working normally
• Another personal account on the same Mac: Working normally

Relevant Diagnostic Logs

Previous Codex Desktop logs showed:

WebSocket was closed before the connection was established
closeCode=1006
SSH: codex path probe timed out after 60000ms
app_server_connection.connection_failed_before_ready
app_server_connection.reconnect_failed

These logs indicate repeated connection failures. However, I cannot confirm whether they directly caused the model capacity error.

Feedback Upload Issue

I attempted to submit in-app feedback including logs.

Feedback ID:

01a12321-cbc7-7e12-8fd5-320a3534087a

Codex Desktop displayed:

Feedback upload not confirmed. We couldn't confirm whether your feedback was received. It may have been delivered, and resubmitting could create a duplicate.

The diagnostic upload therefore remains unconfirmed.

Important Observations

• The problem follows the affected account rather than the computer.
• Switching to another personal account immediately allows Codex Desktop to work normally.
• The affected account has full usage allowance.
• The affected account can use Codex Web normally.
• The problem recovered for approximately one day, then returned.
• The issue affects every available model, not just one model.

Investigation Request

Could the engineering team investigate possible account-specific model access restrictions, request admission or routing problems, desktop authentication issues, or account-associated execution environment failures?

The controlled comparison between two personal accounts suggests that a general device-wide network or installation issue is less likely.

Thank you.