gpt-5.5 is listed as available locally but real requests fail with 404 'Model not found' in both Desktop and CLI
Summary
On June 7, 2026, gpt-5.5 became unusable for me on the same machine in both Codex Desktop and Codex CLI, while gpt-5.4 still works normally.
The key issue is that local model metadata still says gpt-5.5 is available, but actual requests to the Codex responses endpoint fail with:
unexpected status 404 Not Found: Model not found gpt-5.5
This looks like a mismatch between model discovery/model availability metadata and the actual backend that serves Codex responses.
Environment
- OS: Windows
- Auth mode: ChatGPT
- Plan: ChatGPT Plus
- Codex CLI:
0.130.0 - Desktop app build seen in logs:
OpenAI.Codex_26.602.4764.0_x64__2p2nqsd0c76g0 - Local config currently defaults to
gpt-5.4, but forcinggpt-5.5still reproduces the failure
What I verified locally
gpt-5.4works.gpt-5.5fails in Desktop.gpt-5.5also fails in CLI on the same machine/account.~/.codex/models_cache.jsonwas refreshed on2026-06-07T15:47:34Zand still includes:
{
"slug": "gpt-5.5",
"supported_in_api": true
}So the local model list says gpt-5.5 is available, but real requests are rejected.
Minimal CLI repro
This succeeds:
codex.cmd exec -m gpt-5.4 "Reply with exactly OK."This fails:
codex.cmd exec -m gpt-5.5 "Reply with exactly OK."Error:
ERROR: unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a080d04cb836af8b-NRT, request id: 759e8b79-b0b8-4c5d-8ab3-f476b8cdfd87I also saw the same class of error in Desktop with a different request id/cf-ray.
Expected behavior
If gpt-5.5 is shown/listed as available, real requests should succeed.
If gpt-5.5 is no longer available for my account/region/client combination, it should not be advertised by the local model list / model picker.
Actual behavior
- Local model metadata advertises
gpt-5.5 - Actual responses endpoint rejects it with 404
Model not found gpt-5.4still works with the same auth/session/machine
Likely issue
This looks like one of:
- model discovery metadata is stale or over-advertising
gpt-5.5 - account entitlement / rollout state for
gpt-5.5is out of sync with the responses backend - Codex clients are receiving a model list that is not consistent with what
https://chatgpt.com/backend-api/codex/responseswill actually accept
Additional note
This does not look like a generic auth failure, because gpt-5.4 works immediately as a control case on the same machine and account.
86 Comments
Same problem here.
Same issue here on macOS with ChatGPT Plus.
My case matches the core pattern in this issue:
gpt-5.5works in normal ChatGPTgpt-5.4works in Codexgpt-5.5fails in both Codex Desktop and Codex CLI with404 Model not foundDesktop error:
unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, request id: 54446a05-6a0d-49e6-8e2e-28aee034bb33Compaction also fails:
Error running remote compact task: unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses/compact, request id: 48e90786-b13c-45d5-a452-dde67d6e4394CLI also fails on
gpt-5.5, including fallback after agpt-5.5-codex-...alias attempt.I also reset Codex auth and local model cache, and the same 404 persists.
Same problem here.
unexpected status 404 Not Found: Model not found gpt-5.5,
Is it due to issues in the Japanese region？
I encountered this issue 12 hours ago, and I only saw someone mention it on X. It hasn't been fixed until now, probably because Asians are all sleeping
The day before yesterday, there was a false block. Yesterday, the model made an error
What are you up to, OpenAI
In Korea, Same problem here.
⚠ 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-
control
■ unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a08416e9db97560a-ICN,
request id: b8256f4c-ab48-4b41-a963-4574e9fa84dc
Same problem.
unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses,
problem here in Brunei region
+1
same in Singapore
same problem
same in Taiwan, but my system is macOS
same problem
yes,same problem
Same issue here, but not sure if it is JP only; in Mainland China now and switched Proxy to different countries (such as US) still encounting the same issue: "unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses"
same issue here in Malaysia
Same problem.
same in Australia
Same Problem.
"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:
a084a156ff78f8ce-SIN, request id: 564a5340-3ab2-4911-8b87-fdfbe909b74a"
did we just a witness a global downtime of employee's productivity in coding? XD
same in Australia
same in Indonesia
Adding a quick triage summary based on the reports so far:
404 Model not found gpt-5.5behavior.gpt-5.4works, whilegpt-5.5fails.gpt-5.5-codex-...aliases/treatments, which may help narrow this down to model routing / availability metadata rather than a generic client-side problem.So at this point the issue looks more like a backend-side model availability / routing inconsistency than an isolated local configuration problem.
same in macOS
This phenomenon seems to be happening to certain pricing plan customers.
I started the subscription today, but I got this error today, so I refunded it through openAI customer center.
After I got my refund, I tried again a while later and found Chatgpt 5.5 working and I thought it was bug patched, so I subscribed again.
However, the same error occurred again after the subscription.
The pricing plan I use is plus.
I can reproduce a similar issue, with a few additional observations.
After retrying several times,
gpt-5.5can eventually work without changing the model, auth state, account, config, workspace, or local cache.My
/statusshows the public model alias normally:However, some attempts fail with:
For my own machines, I have only observed this problem on my Tokyo machine. I have not observed the same issue on my machines or servers in Taiwan with the same account and model setup.
Another detail: retrying inside the same broken session does not always recover it. In my case, I may need to restart/resume the Codex session with
codex resume, and then it can start working again.Since the same public alias can eventually work, this does not look like a purely local configuration issue or a permanent account entitlement issue. It looks more like a non-deterministic Codex backend routing / model alias propagation problem: the client accepts
gpt-5.5, but some backend requests appear to be routed to an internal deployment name that is not available on every serving path yet.In other words,
/statusand the model picker can showgpt-5.5as valid, but actual inference requests may still intermittently hit a backend where the resolved internal deployment is missing, producing a 404. Retrying or resuming the session can then land on a working route/backend.~~I logged out and use device auth to sign in to fix the problem.~~
After I logged out from codex, I have
Sign-in failed: failed to start login server: An attempt was made to access a socket in a way forbidden by its access permissions. (os error 10013) codex
So I used the CLI device auth to sign in again.
Then the gpt-5.5 Medium can be connected and ready to use.
edit:
the gpt-5.5 medium worked about 3 minutes and back to 404 again.
same issue here in indonesia.
I login use my laptop.. Before, I use my account to login at codex cli for Hermes in vps and my PC at home.. Today I login same account to my laptop for traveling. Then this issue appear.. Does this mean soon we can't use codex with gpt 5.5 on multiple devices?
Mine is plus $20 plan
My friend using pro $100 plan and it's working fine. No error at all
It seems to be back to normal now. Is it working for anyone else?
UTC+8 15:30
macOS
nope. It's still happening on my PC.
Today is Monday, and my work progress is 0, LOL
It worked fine for a while, but then the “404 Not Found” error appeared again.
Just worked for a few minutes and back to failing.
Switch to Claude =)
me neither
It worked fine in last 1.5 hours, then 404 comes back.
Hermes using codex CLI normal now I let it run check with gpt-5.5 model with result 30k context used and worked fine
Windows haven't tried it again since I'm on the road now
Same problem here
me too......................
<img width="729" height="614" alt="Image" src="https://github.com/user-attachments/assets/218cefae-d0f2-43fc-85a0-ff9e639dc56b" /> the audacity to tell there is no issue with codex/chatgpt5.5
有时候可以，有时候不行，很烦。。。。
seems to be an issue with asia region
same issue, but if I use codex auth with opencode, it works fine
<img width="268" height="147" alt="Image" src="https://github.com/user-attachments/assets/ba5fa7a0-8a5c-4c75-bf9a-3e70e8078bab" />
_same issue?_
Worked for 2 minutes after restarting the app and then stopped working again. It's been like this since morning.
unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses,
@tibo-openai , looks like a bunch of us, mostly localized to Asia, aren't able to use 5.5.
Inicie un nuevo Hilo, pareció responder bien, pero luego volvió el mismo error. Escribo desde Argentina.
I started a new thread, initially received a response, but then the same error returned. I'm writing from Argentina.
Facing the same issue, it works intermittently though.
Can anyone confirm their account subscriptions too?
plus $20 got this error 404.
$100 pro account unaffected at all and work normally
Location: Indonesia
Does they plan something on cheaper account lol
Plan Plus error 404
ChatGPT plus
macos
unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a0880acaef0ace4f-SIN, request id: bc43ac5d-ce69-4142-ab7d-4cfbb6a1bf4e
Sri Lanka
Same issue from Brazil:
unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a08831911d7bf1cd-GRU, request id: eac3747d-008a-4295-95c8-0ad22e2586b8
<img width="351" height="61" alt="Image" src="https://github.com/user-attachments/assets/2d829fec-d1dc-47b2-a94a-685da32599e1" />
Same problem
Same issue here.
Environment:
model = "gpt-5.5"
model_reasoning_effort = "xhigh"
Error:
unexpected status 404 Not Found: Model not found gpt-5.5
url: https://chatgpt.com/backend-api/codex/responses
cf-ray: a08857103bea684e-EZE
request id: 71112844-a7b0-4169-915e-4f69fc367f0b
same issue
same issue: unexpected status 404 Not Found: Model not found gpt-5.5
It's been over a day and the issue still hasn't been resolved. Around noon today, Codex worked successfully twice, but after that it started showing the same 404 error again: 'unexpected status 404 Not Found: Model not found gpt-5.5'.
Still same issue here .
■ Error running remote compact task: unexpected status 404 Not Found: Model not found gpt-5.5, url:
https://chatgpt.com/backend-api/codex/responses/compact, cf-ray: a088e10b6ec555f9-KUL, request id: 65319c06-caa7-92bb-9830-7232086ea4e7
same issue here !
Has the official Codex organization noticed this?
same issue here. thank you for posting.
I wasted so much token debugging this unsolvable bug. (sigh)
Same issue here.
Now it works
Sometimes, it works, Sometimes not
It seems like only happened in individual account
, Business account is working
Same problem
Error running remote compact task: unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a0897e42ca2933b6-BKK, request id: eaca235e-84ff-45c6-8a72-7cf276a9541d
Codex 版本 26.602.40724 • 发布于 Jun 5, 2026
Same issue here
Hi everyone! Sorry, we are looking into this. Will report back here shortly.
Same issue here.
Same issue here.
Same issue here.
unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a089e339987a684e-EZE, request id: 2dc27620-e7bd-486b-bec4-c518f4a6e72d
I've been crying non stop because of this
The strongest extra signal here is that this is failing across both Desktop and CLI while the same ChatGPT auth path still serves gpt-5.4. That makes it look less like a generic client-side cache problem and more like a discovery/eligibility mismatch between the advertised model list and what the Codex responses backend will actually accept for this account or rollout bucket.
death before 5.4
I came across same issue, I cannot use gpt-5.5 and it said
unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a08a0d888e6df1d0-KHH, request id: 63c6521d-9ddb-4f4a-bb8c-cfd648b3eb11, but gpt-5.4 is working.This issue is sustianed to right now.
Same issue here-
getting this error- unexpected status 404 Not Found: Model not found gpt-5.5-codex-1p-exp-p-0604-811f2f-ev3-text-controlOther models work fine
It only happened to plus account. I have pro and plus. Pro working fine.. Then few hours ago plus start working fine.. But, The usage drain is like become super fast not like usual.
If some of you also experience this and can normally run again after 404.. try use it and check the usage remaining 5h and weekly.
I don't know if it just me.. but I used it to work like usual and the 5h usage remaining gone by more than half less than ten minutes only 45% left lol
Yeah @michaelxer , the usage seems to drain faster than usual. Is there any thread that discusses about the fix for the last night issue?
same issue here. I use desktop codex and have a plus account. gpt-5.4 works fine. the problem happened the day before yesterday, and when I switched to gpt-5.4, it seems the usage decreases extra fast :(
unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a086053b2e911a63-ICN, request id: 49667fa6-5832-4d3e-862c-6cbaaf23abbb
Thanks for your patience. The backend issue causing
gpt-5.5requests to return404 Model not foundis now resolved.Please try again. If it still fails, share a fresh request ID and timestamp.
Same problem here: unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a1022bf2cdb88df6-MEL, request id: 17c6fbda-ab09-4edc-8e20-4d6d01a396e7
Meet the same problem: unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a102f4082b4cb701-SJC, request id: 0940c4f9-b66c-40ea-b81c-266d0b69487a
I've got the same problem today: unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a1036303c9c5d625-LCA, request id: 302d1939-33b5-4339-804c-867dc24c6499
Same issue here: unexpected status 404 Not Found: Model not found gpt-5.5, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: a103fc223f27dea1-DAR, request id: 65d318bb-ed50-47cf-bb1e-0949a2745815
Mine works after unselecting fast mode
Thank you very much; the issue has been resolved.
For anyone else struggling with this issue. Am using Codex plugin in VS Code.
None of the recommended model and speed setting changes above worked for me. What did work: installing a prior plugin release from 3-4 days ago and restarting the plugin.