unexpected status 404 Not Found: Model not found gpt-5.6-luna ON CODEX APP WINDOWS. GPT 5.6 luna not working IN WINDOWS 11 PRO.

---

Title: gpt-5.6-luna returns 404 Model not found in Codex despite docs showing it as available

Description

I’m seeing a hard failure when trying to connect to gpt-5.6-luna in Codex. The request reaches the Codex backend, but the backend responds with a 404 Model not found, which makes this look like a model availability / entitlement / routing issue rather than a local network problem.

Error

unexpected status 404 Not Found: Model not found gpt-5.6-luna, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a19dc3a64d36c174-BLR, request id: f07b19a8-b6c5-4096-98f1-030227ed55c2

Environment

• Codex desktop app windows 11
• Current date at time of testing: 2026-07-12
• Model attempted: gpt-5.6-luna

What I expected

Codex should be able to route the request to gpt-5.6-luna successfully, since the model is documented as available.

What actually happens

The backend rejects the request with:

• HTTP 404 Not Found
• Model not found gpt-5.6-luna
• Endpoint: https://chatgpt.com/backend-api/codex/responses

What I tried

1. Confirmed this is not a simple network outage.

• The backend is reachable and returns an application-level error rather than timing out or failing to connect.

1. Checked OpenAI/Codex status.

• The service status appeared operational, so this does not look like a broad outage.

1. Checked the official model docs.

• The docs list gpt-5.6-luna as a valid model, which suggests the model name itself is not a typo.

1. Compared against local Codex config/auth state.

• Local Codex config showed a different default model setting.
• Auth state appeared inconsistent/stale relative to the current date, which may point to entitlement caching or account/session state being out of sync.

1. Retried after looking for local session/auth issues.

• The failure persisted, still returning the same 404 from the Codex responses endpoint.

Why I think it is failing

The most likely causes are one of these:

• gpt-5.6-luna is not enabled for the current account/session.
• Codex has stale cached entitlement or auth state.
• There is a backend rollout / routing mismatch where the model is documented but not resolvable for this client surface.
• The model name is valid in docs but not currently available in this specific Codex build or account tier.

Notes

This does not look like a code issue in my repo. The request is failing before any repository logic matters, at the Codex backend model resolution step.

Request

Could you verify whether:

• gpt-5.6-luna is currently enabled for all Codex users / plans,
• there is a known entitlement or rollout issue,
• or the Codex client should be updated to handle this model name differently?