Codex CLI Update: What's New in v0.160 (Oct 2026)

Codex CLI has shipped a quick run of releases in the last two weeks, and a few of them change how you use it day to day. Here is everything you need to know, without digging through the changelog.

What Changed, At a Glance

New to the tool? Start with our Codex CLI setup guide first, then come back here.

What's New in the Latest Codex CLI Versions

Instant interrupt

Added in v0.159.0. When you turn on the instant_interrupt option, you can type new input while Codex is still working and it will use that to steer the response, instead of making you wait or cancel. For long tasks this saves a lot of restarts. It is opt-in, so nothing changes until you enable it.

GPT-6.1 Sol is now the default model

From v0.159.1, GPT-6.1 Sol is the default model in the bundled model catalogs, including Amazon Bedrock. You can still switch models inside a session with /model. If your results feel different this week, this is the first thing to check.

Guardian reviews and command center history

v0.160.0 improves the opt-in Guardian review feature, which can pull in earlier instructions and context when work is handed between agents. It also adds history browsing in the agent command center, plus keyboard navigation and projectless sessions.

Smaller changes worth knowing

• A compact welcome screen and better Mermaid diagram rendering
• Copy-on-select setting and OAuth client secret support for MCP servers
• Stronger security for MCP and WebSocket connections
• Terminal approval is on by default for elevated permissions
• Optional reminders to finish account security setup if you sign in with ChatGPT
• Removed: automatic follow-up prompt suggestions and the bundled plugin-creator skill

Windows Users: What Got Fixed

Two Windows fixes landed back to back. v0.159.2 stops console windows from flashing when Codex starts background processes and sandboxed commands. v0.160.1 fixes the Windows remote MCP environment so it keeps its settings. If you saw odd pop-up windows or broken MCP tools on Windows, update first and test again.

How to Update Codex CLI

If you installed with npm, run:

npm install -g @openai/codex@latest
codex --version

If you used Homebrew or the standalone installer, re-run the same install command you used originally. After updating, codex --version should show 0.160.x or newer.

If something breaks after the update, our guide to Codex database and state errors covers the most common fixes.

Codex Usage Limits by Plan

With a new default model, many people are asking how much they actually get. Per OpenAI's pricing page at the time of writing:

For Plus and standard Business, OpenAI lists these local message estimates per five-hour window:

Limits depend on task size and whether you run locally or in the cloud. Fast mode uses 2.5x of your included limit, and GPT-6 Astra Ultrafast uses 8x. Pricing and limits change often, so check OpenAI's pricing page before you plan around them.

Which Model Should You Use in Codex?

A simple rule of thumb from the numbers above: use the default Sol for everyday coding, move to Astra only for hard problems that justify the smaller message budget, and use Luna for quick, cheap edits where you will send many messages.

Want to see what Codex can do once it is set up? Read how Codex can replace routine junior developer work, or check out the new record and replay feature.

FAQs