OpenAI published Codex CLI 0.166.0 on GitHub as tag rust-v0.166.0 at 2026-10-29T19:47:53Z (2026-10-30 03:47 Asia/Shanghai). The full changelog compares against rust-v0.165.0. Related reading: Codex CLI 0.165.0 Adds Vertex AI Provider and Memory v2 Defaults. The usual publish path is npm i -g @openai/codex@latest (npm latest listed 0.166.0).
Day-to-day CLI changes
The GitHub Release highlights keyboard-driven transcript selection (word／line／page movement, search, and configurable shortcuts). Async Q&A／choices／answers／transcript replies gain consistent Markdown rendering. codex queue --resume lets a queue message resume a thread that was not loaded. TUI safety-block messaging is clearer, with an optional safer follow-up path. codex doctor --network-only focuses diagnostics on configuration, connectivity, and WebSocket checks.
GPT-6.1 Terra was added to the bundled／Vertex AI／Amazon Bedrock model catalogs only—a Codex catalog entry in this cut, not a separate model launch announcement here. 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. Coding-agent contrast: Claude Code vs Codex.
Bug fixes (brief)
Release notes also call out resume／fork keeping approval and permission state; composer image-attachment fixes; macOS standalone cleanup for non-ASCII paths; unarchive of child threads; app-server start／stop races; TUI reconnect／offline /quit behavior; and resume／fork preserving reasoning-token accounting so compaction is less likely to fire early.
Chores in the same tag (Bazel／V8, standalone executor packaging, Node.js 24.22.1／Rust 1.95.0) are build／toolchain work, not the product headline.
Product details can change; the latest GitHub Release wording applies. For reference only.