gpt-5.5 returns 404 Model not found while gpt-5.4 works in Codex App/CLI

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

Version 26.616.71553

CLI bundled with the app:

codex-cli 0.142.0

What subscription do you have?

ChatGPT Plus

What platform is your computer?

Darwin 25.5.0 arm64 arm

What issue are you seeing?

gpt-5.5 is currently failing in Codex with 404 Model not found, while gpt-5.4 works from the same machine/account/auth path.

The Codex app initially retries WebSocket, falls back to HTTPS, then fails against the Codex responses endpoint.

Observed minimal repro error:

warning: Falling back from WebSockets to HTTPS transport. unexpected status 404 Not Found: Model not found gpt-5.5-codex-premium-1p-exp-p-0616-3a9cde-ev3-text-2-treatment
ERROR: unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a10060497925480c-KIX, request id: a639b1bd-d3c1-4324-9b97-b28f09d9aee6

A prior app-thread failure showed the same class of error:

unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a100562a8bdb8372-KIX, request id: e0d6b04d-89dd-4c64-8d18-e8e0c4f21a887

This looks like a recurrence of the earlier backend/model-routing issue from #26892 / #26902 / #27010, but with a newer internal treatment suffix 0616.

What steps can reproduce the bug?

From the same host/account:

/Applications/Codex.app/Contents/Resources/codex exec -m gpt-5.5 -s read-only --skip-git-repo-check --ephemeral 'Reply with exactly: OK'

Result: fails after reconnecting 5/5 and returns 404 Model not found gpt-5.5.

Control case:

/Applications/Codex.app/Contents/Resources/codex exec -m gpt-5.4 -s read-only --skip-git-repo-check --ephemeral 'Reply with exactly: OK'

Result: succeeds and returns OK.

codex doctor reports config/auth/state/connectivity as healthy, so this does not appear to be local config or SQLite corruption:

Configuration: config loaded; model gpt-5.5; auth configured; stored auth mode chatgpt
State: databases healthy
Connectivity: websocket connected (HTTP 101 Switching Protocols)
Reachability: ChatGPT base URL reachable

Session id from the minimal failing CLI repro:

019ef281-07ca-70a0-b24f-773209ca9b2c

What is the expected behavior?

If gpt-5.5 is listed/selected as available in Codex, requests should succeed.

If this account/region/treatment cannot currently use gpt-5.5, the model picker / local model metadata should not advertise it as available, or the error should identify the entitlement/routing problem instead of returning 404 Model not found.

Additional information

Local model cache includes gpt-5.5 and the official Codex models documentation lists gpt-5.5 as a recommended Codex model. The failure appears specific to the backend route for gpt-5.5; gpt-5.4 works as a control on the same machine and account.