Skip to main content

How to Find the Current ChatGPT and Codex App Build on macOS

Use OpenAI's official app feeds to find the current macOS stable or beta build, verify the version, and avoid stale direct ZIP links.

By

OpenAI launched the Codex desktop app for macOS on February 2, 2026. It has since become part of the unified ChatGPT desktop app: you install one app, then choose ChatGPT or Codex for the work you are doing.

The underlying update feeds still use Codex-era URLs. They remain the durable way to check the current macOS build. Direct ZIP links are useful when you need the file immediately, but they age.

If you are looking for Windows instead, use this companion post:

ChatGPT Desktop with Codex on Windows

For Linux, the desktop app path is currently unofficial:

How to Install ChatGPT Desktop with Codex on Linux

Current Snapshot

Checked on August 11, 2026:

ChannelCurrent feed entryPublishedMinimum macOS
Stable26.803.61601Mon, 10 Aug 2026 18:19:35 +000012.0
Beta26.727.40816Thu, 30 Jul 2026 19:38:05 +000012.0

The stable channel is currently newer than the beta channel. Treat them as separate release streams rather than assuming beta always has the higher version.

Use the Feed, Not a Screenshot

The stable macOS feed is:

https://persistent.oaistatic.com/codex-app-prod/appcast.xml

The beta macOS feed is:

https://persistent.oaistatic.com/codex-app-beta/appcast.xml

Open the feed and read the first <item>. That top item is the latest build for that channel.

The fields that matter are:

  • <title> for the public version
  • <pubDate> for when the feed entry was published
  • <sparkle:minimumSystemVersion> for the minimum macOS version
  • enclosure url for the direct ZIP download

As of this update, the current Apple Silicon links are:

Use those links if you need this exact build. Share the feed if you want the link to stay useful after the next release.

Stable vs Beta

The stable channel is the normal ChatGPT desktop app. The beta channel is useful when you are deliberately testing the beta app or checking whether a reported beta-only change has shipped.

Do not mix the two when documenting a bug, release, or install path. A stable build and a beta build can share a version number while still installing as different apps.

OpenAI’s current desktop documentation points to the unified ChatGPT app for macOS and Windows. If you are installing for daily use, start there first:

OpenAI desktop app docs

GitHub Releases Are a Different Stream

The openai/codex GitHub releases page is not the desktop app update feed.

GitHub releases track the open-source Codex project and CLI release stream. The macOS desktop app is distributed through OpenAI’s appcast feeds. If the GitHub version and the desktop version look unrelated, that is expected.

How to Install on macOS

  1. Choose the stable or beta feed.
  2. Open the feed and use the first item.
  3. Download the ZIP from the enclosure url.
  4. Extract the app bundle.
  5. Move the app into /Applications.
  6. Open it, sign in with your ChatGPT account, and choose Codex when you want the coding workspace.

If macOS shows the normal downloaded-from-the-internet prompt, review it and continue only if the file came from the OpenAI feed.

Terminal Check

To print the current stable ZIP:

Terminal window
curl -fsSL https://persistent.oaistatic.com/codex-app-prod/appcast.xml | grep -m1 -o 'https://[^"]*\.zip'

To print the current beta ZIP:

Terminal window
curl -fsSL https://persistent.oaistatic.com/codex-app-beta/appcast.xml | grep -m1 -o 'https://[^"]*\.zip'

That command does not prove the app is right for your machine. It only extracts the newest direct URL from the relevant feed.

An older beta link that circulated in March was:

Codex (Beta)-darwin-arm64-26.311.30926.zip

That file may still resolve, but it is not the current beta build. Compare the version in any shared filename with the first item in the relevant appcast before reposting it.

Sources Checked

About the author

Ashton

Systems Engineer and IT consultant working across infrastructure, healthcare IT, automation, and practical security.