bug(cli): gpt-5.5 returns 404 Model not found while gpt-5.4 works
Resolved 💬 3 comments Opened Jun 8, 2026 by estudiochapunov Closed Jun 12, 2026
What version of Codex are you using?
codex-cli 0.137.0
Which model were you using?
gpt-5.5
What operating system are you on?
Linux x86_64
What issue are you seeing?
Using Codex CLI with a ChatGPT account, gpt-5.5 is currently unusable while gpt-5.4 works on the same machine/account.
Observed behavior:
- Starting Codex with
codexand defaulting or switching togpt-5.5fails. gpt-5.4works from the same environment.- Codex had been updated to the latest version roughly 30 minutes before reproduction.
Representative errors:
⚠ Falling back from WebSockets to HTTPS transport. unexpected status 404 Not Found: Model not found gpt-5.5-codex-1p-exp-p-0604-811f2f-ev3-text-1-treatment
unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a0884ce7586bca16-EZE, request id: db9ace7c-1ace-4d1a-8d2b-d6f7e82aaf6c
Model changed to gpt-5.5 low
unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a088508a598614a0-EZE, request id: ba2370ef-e800-43c1-aeaf-b34b2204bd49
This does not look like a generic local-network failure because another model (gpt-5.4) works in the same session/account.
Steps to reproduce
- Update Codex CLI to the latest available version.
- Run
codexorcodex --model gpt-5.5. - Sign in with a ChatGPT account if needed.
- Send any prompt, or switch models to
gpt-5.5from within Codex.
Expected behavior
gpt-5.5 should accept requests when selectable/available, or Codex should return a clearer entitlement/availability message instead of a backend 404 Model not found.
Actual behavior
Requests to gpt-5.5 fail with backend 404 Not Found: Model not found gpt-5.5, while gpt-5.4 remains usable.
Additional context
- This resembles the earlier April 2026 Codex incidents involving intermittent
not founderrors forgpt-5.5. - Public status currently shows Codex operational, so this may be a partial rollout / entitlement / catalog mismatch rather than a fully declared outage.
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗