🤖 fix: cap GPT-5.6 context over Codex OAuth#3724
Conversation
|
@codex review |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 6486e29d00
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
coadler
force-pushed
the
fix/gpt-56-oauth-context-cap
branch
from
6486e29 to
b5e3c51
Compare
|
@codex review Updated the OAuth overrides to use 272K for Sol/the bare alias and 128K for Terra/Luna, with tier-specific test coverage. |
|
Codex Review: Didn't find any major issues. Chef's kiss. Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
The CODEX_OAUTH_CONTEXT_WINDOW_OVERRIDES doc comment already explains that these caps are kept separate from model metadata so API-key requests retain the public window. #3724 rewrote the inline comment to restate the same point, so trim the duplicated sentence and keep only the tier-specific rationale. Comment-only; behavior-preserving.
1 participant
Summary
Apply the ChatGPT/Codex OAuth context caps to the GPT-5.6 family while retaining the public 1.05M context window for API-key requests. Sol and the bare GPT-5.6 alias use 272K; Terra and Luna use 128K.
Background
The ChatGPT/Codex OAuth routing layer has lower context limits than the public OpenAI API. GPT-5.6 was initially allowed through OAuth without route-specific context overrides, causing compaction and the token meter to use the API's 1.05M limit.
Implementation
Risks
Low. The overrides only apply when direct OpenAI requests actually route through Codex OAuth. API-key and gateway-routed requests continue using their existing model metadata.
Generated with
mux• Model:openai:gpt-5.6-sol• Thinking:medium• Cost:$5.63