Codex Desktop 26.623.101652 shows selected model capacity error despite available usage

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

Version 26.623.101652, build 4674

Local CLI version: codex-cli 0.142.5

What subscription do you have?

Codex access is enabled. The exact public plan name is not included in this report.

What platform is your computer?

macOS 26.3.1, arm64

What issue are you seeing?

After the recent Codex Desktop update, sending or continuing a prompt can immediately show this banner:

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

The selected model shown in the UI is GPT-5.5 with medium reasoning. The app still appears usable and the access/model controls are visible; the UI does not indicate that the normal 5-hour or weekly usage quota is exhausted.

This looks like a current desktop build capacity / rate-limit / entitlement mismatch, rather than a local app startup failure.

What steps can reproduce the bug?

I do not yet have a deterministic minimal reproduction.

Observed flow:

1. Open Codex Desktop 26.623.101652.
2. Use GPT-5.5 with medium reasoning.
3. Enter a prompt in an existing conversation.
4. Attempt to send or continue the conversation.
5. Codex shows Selected model is at capacity. Please try a different model. before the request can proceed.

What is the expected behavior?

If the selected model is temporarily unavailable, Codex should clearly distinguish model capacity from account quota or usage-limit exhaustion.

If the account still has usage available, the app should either allow a retry/fallback path or explain which specific model, reasoning tier, entitlement bucket, or temporary service condition is blocking the request.

Additional information

A screenshot was captured locally showing the banner and the GPT-5.5 medium model controls. The UI around the composer is localized in Korean, so the relevant visible banner text is copied above.

This report is for the current desktop build and is separate from the older June capacity incident reports.