Model routing mismatch: requesting gpt-5.3-codex returns gpt-5.2-2025-12-11 (verified via SSE response.created)
Resolved 💬 19 comments Opened Feb 15, 2026 by kwikysw Closed Feb 18, 2026
💡 Likely answer: A maintainer (github-actions[bot], contributor)
responded on this thread — see the highlighted reply below.
What version of the Codex App are you using (From “About Codex” dialog)?
Version 26.212.1823 (661)
What subscription do you have?
ChatGPT Pro
What platform is your computer?
_No response_
What issue are you seeing?
Summary
When I request gpt-5.3-codex, the backend creates the response with response.model = gpt-5.2-2025-12-11 (verified via SSE response.created). This looks like a silent downgrade and matches the “feels like 5.2 / slower / less progress updates” behavior in the Codex app.
Environment
- codex-cli: 0.93.0
- Codex desktop app version: <fill in from “About Codex”>
- Subscription: Pro
- OS: macOS <fill in version>
Steps to reproduce (CLI, produces debuggable proof)
Run:
RUST_LOG='codex_api::sse::responses=trace' \
codex exec --skip-git-repo-check -m gpt-5.3-codex "Reply with exactly: PING" 2>&1 \
| python3 -c 'import json,re,sys
for line in sys.stdin:
m=re.search(r"SSE event: (\{.*\})", line)
if not m: continue
evt=json.loads(m.group(1))
if evt.get("type")=="response.created":
r=evt["response"]
print("response.model:", r.get("model"))
print("response.id:", r.get("id"))
break'
Expected
response.model should reflect gpt-5.3-codex (or a gpt-5.3-* snapshot).
If a fallback is required, it should not be silent (UI/CLI should clearly indicate the served model).
Actual
The response.created event reports:
response.model: gpt-5.2-2025-12-11
response.id: resp_0c6ab9820949f9f10169922a3449c88191afe5e6a021ad74cf
Impact
Noticeably worse/slower behavior when selecting/using “5.3” compared to prior 5.3 behavior, consistent with being routed to 5.2.
Additional context
Observed on: 2026-02-15 (local time)
Possibly related: #11189, #11187, #11215 I know that this is a "duplicate". This isn't normal. Not alerting people about getting a worse model than what you paid for, staying silent about it too, will you refund requests that were done with gpt 5.2 at least ?
What steps can reproduce the bug?
Do a request with GPT 5.3 codex and check sse logs to see what model was used as described.
What is the expected behavior?
For requests that are done with GPT 5.3 codex selected to actually use GPT 5.3 codex.
Additional information
_No response_
19 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
You've been silently downgraded and need to send your Passport pic to OpenAI
What a joke
<img width="699" height="1200" alt="Image" src="https://github.com/user-attachments/assets/af190081-9a55-450d-8450-702058923ba3" />
Yep I know that and I am NOT sending my ID to use a freaking coding model !
Getting the same thing and I'm verified
Getting the same thing and im verified
Sending your ID doesn’t fully solve it btw. 48 hours after verification, I was downgraded to GPT‑5.2 again, so their classification is broken, and if you get downgraded you'll be stuck there for sometime.
I’m being incorrectly routed to GPT-5.2. I'm pro sub and cyber verified
Thanks for reporting, and apologies for the inconvenience. We're aware of the issue and are actively working on a mitigation. I'll share updates as they become available.
Thank you Eric. Not that you are alien to knowing how to redeem the situation, but I would expect rate limit resets for the amount of tokens we spent on the wrong model.
Also Please make the new status line something that take the model being used from the response api return and not be fake from the options.
Best,
This is extremely disappointing. I'm doing legitimate work and I've been suddenly downgraded, only discovering after seeing this happen to multiple people on Twitter and then checking for myself. I've performed the cyber verification and I'm still being downgraded across my entire account.
This is outrageous. Codex's advice as a workaround is to basically carefully word my prompts such that they don't trigger the classification model's guardrails. What the hell is this about? What do you think an actual black hat is going to do? You're just frustrating genuine paying users from getting the access that they've paid for.
Are users who are red-teaming their own product going to be permanently rerouted to a worse model? How would OpenAI accurately determine whether a user’s prompt is not maliciously targeting unauthorized third parties, especially at scale?
Users who have already completed the chatgpt.com/cyber verification are still being rerouted. In my case, after verifying and using Codex for an hour, I returned to chatgpt.com/cyber only to see:
Clicking “Re-verify” does nothing.
Codex is entirely unusable at this point. If something as simple as selecting a model cannot be relied upon to work consistently, how is a user supposed to trust the rest of the application?
Ironically, the only thing this “Trusted Access for Cyber” rollout has accomplished is sowing distrust on both sides.
I'm back on 5.3 again since I sent a bug feedback in the app. hopefully it stays like that
@etraut-openai Can we have an update here? OpenAI’s Terms of Use do not provide for arbitrary model downgrades of verified paid users outside of billing failure scenarios. This conduct—routing paying subscribers to an inferior model tier while charging for the higher one—likely constitutes an unfair and deceptive practice under Section 5 of the FTC Act (15 U.S.C. § 45), as well as violations of California’s Unfair Competition Law (Bus. & Prof. Code § 17200) and Consumers Legal Remedies Act (Civ. Code § 1770). OpenAI is exposing itself to significant state and federal legal liability by continuing this practice. You should not have shipped something so broken. This is extremely disappointing.
after 0.102 you can see clearly: ⚠ Your account was flagged for potentially high-risk cyber activity and this request was routed to gpt-5.2 as a fallback. To regain access to gpt-5.3-codex, apply for trusted access:
https://chatgpt.com/cyber or learn more: https://developers.openai.com/codex/concepts/cyber-safety.
I'm already verified . @etraut-openai
This is insanely frustrating. What’s the point of being verified if you’re still blocked
What’s the point of the block anyway, if the model filters itself on user requests it doesn’t like
It's back
If you experienced a downgrade from
gpt-5.3-codextogpt-5.2despite completing Trusted Access verification, access togpt-5.3-codexshould now be restored. This fix took longer than expected. The team needed to work through some deployment issues over the past day. The fix is now rolling out across our clusters.If you’ve completed Trusted Access verification and are still seeing a downgrade, please open a new bug report and provide details. Likewise, if you're having trouble with the verification process, please open a bug report.
This issue is related to new cybersecurity safeguards we’re implementing for
gpt-5.3-codexand future models. We’ve published additional documentation explaining these safeguards and why they’re necessary.We’re also rolling out client updates that provide clear messaging when a downgrade occurs. The latest CLI release (v0.102.0) includes this notification. Updated versions of the Codex IDE extension and app will be available shortly with the same messaging. The notification links directly to the new documentation and verification instructions.
We recognize that this rollout caused frustration. We’ll learn from this and aim for smoother releases going forward. Thank you for your patience while we worked through this.
same problem here. this is bullshit paying 230 eur per month and you just block access. tried the verification now 3 times and its stated that the account could not be verified at this time