OpenAI published Codex CLI 0.164.0 on GitHub as tag rust-v0.164.0 at 2026-10-23T04:48:10Z (about 12:48 Beijing time on October 23, 2026). The npm package @openai/codex reached 0.164.0 as latest about four minutes later (registry time 2026-10-23T04:52:29Z, about 12:52 Beijing time). The release compares against rust-v0.163.0. We have not installed or exercised this build; the notes below follow the official release body only.
What the release lists as new for day-to-day CLI use: save reusable prompts and slash commands in ten numbered /presets slots, with a keyboard menu for quick access; set model, reasoning effort, speed, and permission profiles directly with inline /model and /permissions arguments; preview viewed and generated images in the fullscreen transcript on terminals that support Kitty or Sixel graphics; authenticate enterprise MCP connections with eligible Codex account grants or a dedicated sign-in, including OAuth clients that require secrets; mask credentials during supported remote Bash commands and Zsh snapshot sessions (including credentials from custom providers); and enable the Amazon Bedrock setup wizard by default, grouping API key entry with other AWS sign-in methods when credentials are detected.
Bug-fix themes called out in the same notes (short form): failed compaction no longer leaves partial summaries, and background compaction failures fall back without panicking; session resume finds compressed histories and restores saved thread settings; Windows HTTPS/WebSocket paths handle proxy authentication challenges with the signed-in user’s Negotiate or NTLM credentials; CLI -c / --config overrides keep working across nested subcommands; queued prompts wait for pending permission selections; and sandbox checks were hardened for case-variant metadata paths on macOS and replaceable symlinks used to locate executables. Chores mention pruning unused installs and bounding daemon logs—useful for disk pressure, not a product headline.
This article is a developer-facing changelog brief for CLI 0.164.0. It is not a patch to DigIntoAI’s Codex usage-limits evergreen, and it is not a Day 18 quota-reset notice. We have not verified any of the listed behaviors on a local install. For adjacent Codex / coding-agent context on the site, see Claude Code vs Codex: what $20 gets you, Codex Dots: delegate and follow up from ChatGPT, and Framer MCP connecting Claude, Codex, and Cursor.
