GPT-5.3-Codex being routed to GPT-5.2
What version of Codex CLI is running?
0.101.0
What subscription do you have?
pro
Which model were you using?
gpt-5.3-codex
What platform is your computer?
wsl in window 11
What terminal emulator and version are you using (if applicable)?
wsl
What issue are you seeing?
Although the CLI displays that gpt-5.3-codex is being used, inspecting the actual SSE response stream shows that the request is routed to gpt-5.2.
There is no warning or fallback notice indicating that a different model version is being used internally.
This discrepancy raises serious transparency concerns. If a user explicitly selects a specific model version but a different model is silently used instead, this can reasonably be perceived as misleading behavior.
What steps can reproduce the bug?
- Run Codex CLI v0.101.0 on WSL (Windows 11).
- Execute the following command:
RUST_LOG='codex_api::sse::responses=trace' codex exec --sandbox read-only --model gpt-5.3-codex 'ping' 2>&1 \
| grep -m1 'SSE event: {"type":"response.created"' \
| sed 's/^.*SSE event: //' \
| jq -r '.response.model'
- Observe the output.
Expected selected model: gpt-5.3-codex
Actual returned model in SSE event: gpt-5.2
What is the expected behavior?
If the CLI specifies gpt-5.3-codex, the backend should execute the request using that exact model.
If fallback or routing to another model (e.g., gpt-5.2) is necessary due to availability, policy, or infrastructure reasons, the CLI or API response should explicitly indicate that fallback occurred.
Model substitution without disclosure should not happen.
Additional information
I am a Pro subscriber and have completed ChatGPT identity verification (Cyber verification).
I have not used any illegal prompts, policy-violating content, or restricted operations. The reproduction steps involve a simple "ping" request in read-only sandbox mode.
Given that I am a verified, paying user who explicitly selected gpt-5.3-codex, I reasonably expect that the chosen model is actually being used.
If this behavior is intentional (e.g., internal aliasing or automatic routing), it should be clearly documented. Otherwise, silent model substitution risks being interpreted as misleading to verified and paying users who rely on explicit model selection.
15 Comments
yeah, I was never able to use 5.3 codex basically, despite being a pro user and did the cyber
what's up
Original thread: https://github.com/openai/codex/issues/11189
Round 2: https://github.com/openai/codex/issues/11561
After seeing those test snippets shared on X, I tested it myself and it turns out I was downgraded too. No notifications, no errors. Never did anything cyber related. Did the cyber verification and am now back on gpt-5.3-codex. Pro plan if it matters, always using latest codex cli on Mac. Don't know when I was first downgraded.
Unfortunately now my trust in the tool is broken. Having to continuously remember to run a shell script to verify that I'm actually getting served the correct model is ridiculous.
@etraut-openai @ae-openai Any news here? Obviously your previous "fixes" are not working.
maybe we gotta sue cause, you know, false advertising, ui says you're using one thing while you aren't
You've been silently downgraded & need to send a passport pic to OpenAI to get 5.3 back.
<img width="699" height="1200" alt="Image" src="https://github.com/user-attachments/assets/b13eadae-4293-4d82-b233-46d3cd28e788" />
This shit still broken
still happening and im verified
yep same here, but unsure if verification is only with my chatGPT account or it also applies to my developer portal API access
as i only use codex with API key not ChatGPT Login
Edit: ya its locked to ChatGPT Plan/login. those using API key, prepaid are locked out.
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.
If you went through our Trusted Access verification and were being rerouted to 5.2, please check if this is still occurring. We just rolled out a fix, and I'd like to get an external confirmation of the fix before I make a broader announcement.
fixed for me
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.
Not surprising.
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
What the fuck. This is shady as fuck!..