feature
Enabled real-time reasoning effort adjustment during conversations
learn.chatgpt.comagent
Users can now change reasoning effort settings for supported models during active conversations instead of only at the start. This provides more flexibility for optimizing model behavior on-the-fly.
Stacks this affects
Codex ships in 1 tested stack here.
More Codex
releaseAdded GPT-6.1 Terra model to the Codex model catalogsfeatureEnabled memory v2 by default for new Codex profilesfeatureAdded built-in Google Cloud Vertex AI provider to Codex CLIfeatureAdded experimental attached thread forks to app-server APIfeatureCodex CLI 0.164.0 adds inline arguments for model, reasoning effort, speed, and permission configuration