-
Notifications
You must be signed in to change notification settings - Fork 8k
Open
Labels
CLIIssues related to the Codex CLIIssues related to the Codex CLIbugSomething isn't workingSomething isn't workingmodel-behaviorIssues related to behaviors exhibited by the modelIssues related to behaviors exhibited by the model
Description
What issue are you seeing?
After completing identity verification, I had normal gpt-5.3-codex access for two days. Then, without any change on my end, I was silently downgraded to gpt-5.2 not even gpt-5.2-codex. The downgrade has persisted for 22+ hours. No error, no warning; the CLI just quietly serves a different model.
Confirmed with:
RUST_LOG='codex_api::sse::responses=trace' codex exec
--skip-git-repo-check -s read-only -m 'gpt-5.3-codex'
'hi' 2>&1 >/dev/null | rg -o --replace '$1' '"model":"([^"]+)"' | head -n1
Output: gpt-5.2-2025-12-11
Expected: gpt-5.3-codex
What steps can reproduce the bug?
- Be an ID verified ChatGPT Pro subscriber.
- Use Codex normally on non-sensitive repos.
- At some point, get silently downgraded.
- Run the trace command above to confirm.
What is the expected behavior?
A verified Pro subscriber requesting gpt-5.3-codex should receive gpt-5.3-codex. If a downgrade happens, the CLI should surface the reason explicitly.
Additional information
- Affects multiple verified Pro users.
- The downgrade is sticky and doesn't self-resolve.
- If this is a safety classifier issue it needs tuning and users need transparency, not silent downgrades.
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
CLIIssues related to the Codex CLIIssues related to the Codex CLIbugSomething isn't workingSomething isn't workingmodel-behaviorIssues related to behaviors exhibited by the modelIssues related to behaviors exhibited by the model