OpenAI published Codex CLI 0.165.0 on GitHub as tag rust-v0.165.0 at 2026-10-27T20:09:38Z (2026-10-28 04:09 Asia/Shanghai). The full changelog compares against rust-v0.164.0. Related reading: Codex CLI 0.164.0 Adds Ten /presets Slots and Inline /model Switches. The usual publish path is npm i -g @openai/codex@latest (npm latest listed 0.165.0 at write-up time).
Day-to-day CLI changes
The GitHub Release notes a built-in Google Cloud Vertex AI provider configured with model_provider = "gcp". New or fresh profiles turn Memory v2 on by default and can import external agent memories. The session picker gains a repository filter, and the agent command center adds directory grouping so sessions across linked worktrees are easier to navigate. For supported models, changing reasoning effort mid-conversation is allowed by default.
GPT-6.1 Terra was added to the model catalogs only—a Codex catalog entry in this cut, not a separate model launch announcement here. Related reading: GPT-6.1 Terra Is Generally Available in GitHub Copilot. For Codex caps and ChatGPT-plan context on the site, see OpenAI Still Publishes No Weekly Codex Caps, So Here's What Each ChatGPT Plan Gets in October 2026.
Experimental attached thread forks appear on the app-server API; that is secondary surface area, not the headline for this CLI release.
Bug fixes (brief)
The same notes call out fresh TUI sessions honoring the server’s default approval reviewer; keeping TUI preferences editable when execution settings are invalid; cancelling remote MCP tool requests after local response futures drop; preventing queued agent messages from restarting work after an interrupt or a switch to Plan mode; stopping empty retries when a provider reports model-not-found; and closing a marketplace URL validation gap that could let Git connect to a host other than the one checked against allow patterns.
This build has not been installed or exercised on this site. Product details can change; the latest GitHub Release wording applies. For reference only.