Codex app: Every prompt fails with "Selected model is at capacity" across multiple models, including new conversations #46190

What issue are you seeing?

Summary

Codex is currently unusable for me. Every prompt I have tried returns the following error, including a simple greeting in a brand-new conversation:

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

I have already tried switching to other available models, but the same error persists. The recovery action suggested by the error message therefore does not resolve the problem.

Incident reference ID

01a07122-f89e-7363-8fb6-97ad44f91e96

I have already submitted diagnostic/feedback materials related to this incident. Please use the reference ID above to help correlate this report with the submitted information.

Detailed symptoms

• A brand-new conversation fails even when the message contains only a greeting. The original test used the Chinese equivalent of "Hello."
• The prompt does not request file access, code execution, or project modifications.
• Switching between multiple available models produces the same error.
• No successful assistant response is returned.
• In the attached screenshot, the interface still displays a "Thinking" status and an active stop button after showing the error banner.
• The usage indicator showed approximately 97% remaining when checked. I do not know whether that indicator covers every applicable limit, but the displayed error does not identify a specific exhausted limit or provide a reset time.

Environment and context

• Application: Codex desktop app.
• Account plan: ChatGPT Pro.
• Date reported: September 17, 2026.
• The issue was initially encountered on Windows.
• The model selected in the supplied error screenshot is labeled "GPT-6 Astra." Other available models were also tried with the same result.
• The exact app version has not yet been recorded.

Troubleshooting already completed

1. Tested a new conversation with a minimal greeting.
2. Switched to multiple other available models.
3. Checked the usage indicator.
4. Submitted diagnostic/feedback materials.

All tested prompts and model selections continued to fail during this incident.

Expected behavior

Codex should return a response when using an available model.

If the request cannot be processed, the app should provide an actionable explanation that distinguishes temporary model capacity from an account usage limit, authentication problem, or another service/client error. If a retry or reset time is available, it should be displayed.

Impact

This blocks all Codex-assisted development work for me. I cannot obtain a response even for a minimal conversational test, so I cannot proceed with my existing projects.

Investigation request

Please investigate the submitted diagnostics and help determine whether this is a service-side capacity/routing problem, an account-specific issue, or a client/session problem.

The current error repeatedly recommends switching models, but every model I have tested fails in the same way.

What steps can reproduce the bug?

1. Open the Codex desktop app while signed into the affected account.
2. Create a brand-new conversation.
3. Select an available model.
4. Send only a minimal greeting. My original test message was “你好” (“Hello”).
5. Observe the following error instead of an assistant response:
   “Selected model is at capacity. Please try a different model.”
6. Switch to another available model.
7. Create another new conversation and send the same minimal greeting.
8. Observe the same error again.

Actual result

Every prompt and model selection tested during this incident failed with the same capacity error. No successful assistant response was returned.

Expected result

Codex returns a normal response, or provides an accurate and actionable explanation of the restriction preventing the request.

Incident reference ID

01a07122-f89e-7363-8fb6-97ad44f91e96

What is the expected behavior?

No response

Additional information