Blog / Developer tools

Codex Model Retirement Oct 14: CLI 0.163 Upgrade and Config Checks

2026-10-18cdnping editors4 min read

The Codex model retirement affects people who sign in with a ChatGPT account. According to OpenAI's changelog, at noon Pacific Daylight Time on October 14, 2026 (UTC 19:00), an older GPT model was retired from Codex. Sessions that use an API key are not affected. If your config or scripts hard-code the old model name, change it to the new default model.

The information below is current as of October 2026 and comes from OpenAI's Codex changelog (linked at the end). This article does not name specific model versions; check the official pages when you need them.

#What exactly was retired on October 14?

The official changelog says that starting at noon PDT on October 14, 2026, the old model is retired for Codex users who sign in with ChatGPT. The OpenAI API and Codex sessions that use an API key are not included. OpenAI recommends switching to the new default model, where your plan and client support it, and updating workspace defaults, saved settings, and any scripts that select the old model.

The new model did not become the CLI default only now. According to the official changelog, it became the default when it launched in late September. So if you never hard-coded a model name, you probably noticed nothing; only people who pinned the old name will run into problems after October 14.

#Am I affected? Check first

Places to check:

1. Look in ~/.codex/config.toml for a model = "..." setting or similar. The exact key name is defined in the official config docs.
2. Search your projects and CI for the old model name, for example grep -rn "old-model-name" ., including shell scripts, codex exec calls, and workspace config.
3. If your team distributes shared defaults, ask the admin to update them too.
4. After changing it, run a simple codex exec task to confirm it returns normally.

If you still get errors after the change, don't blame your config right away. See our earlier write-up, October Codex errors and "Selected model is at capacity": incident log and troubleshooting, to tell a model that is unavailable apart from a capacity problem. For quota questions, see How Codex usage limits work.

#What's new in CLI 0.163?

Codex CLI 0.163.0 was released on October 15. The install command given officially is npm install -g @openai/codex@0.163.0. The new features listed in the changelog:

• A searchable /config panel for viewing and changing preferences.
• You can send a follow-up while an answer is in progress to steer it (steer).
• A configurable leader key in /keymap, defaulting to Ctrl+X, with voice mute moved to F9.
• Managed background services can be restarted from /daemon.
• Independent read-only tools can run in parallel.

On the fix side, the changelog lists the network proxy's DNS and request-header checks, plus paste handling on Windows and WSL. See the changelog for the complete list.

If you have your own keybinding habits, note that Ctrl+X is now the default leader key and voice mute has moved.

#Does the October 8 Ultrafast mode matter to me?

On October 8, OpenAI introduced Ultrafast mode, which it says speeds up generation for the new default model. It is offered to the Pro $500 tier and eligible Enterprise and Edu plans; on Enterprise it is off by default and a workspace owner has to turn it on. Regular Plus users are not included. Check the official page for exact eligibility.

#How can I learn about default-model changes sooner?

Changes like this usually show up in the official changelog before they show up in your terminal. A few habits:

1. Read the changelog before upgrading, looking for "default model," "retired," and "config."
2. Avoid hard-coding model names in scripts. If you must, keep the name in one variable and manage it in one place.
3. Follow cdnping's official site changes (/en/changes), which monitors the OpenAI and Codex changelogs and records additions and removals line by line, with RSS.
4. When you hit errors, check OpenAI status and Is Codex down to rule out a server-side problem first. For a more systematic approach, see How to find out about LLM price and model changes right away.

#Other ways to use Codex without a ChatGPT login

Codex can connect through an endpoint URL plus an API key. For setup, see Custom API endpoints for Claude Code, Codex, and OpenCode. AGI Plan is a commercial partner of cdnping. It offers a step-by-step Codex setup guide that needs no ChatGPT account and bills in Chinese yuan (RMB); see the AGI Plan site. The site states that it is not an officially authorized channel, so evaluate it yourself before choosing.

#References

• OpenAI Codex changelog (now redirects to the learn.chatgpt.com changelog), accessed October 2026