Package Details: chatgpt-desktop 26.930.21537-1

Git Clone URL: https://aur.archlinux.org/chatgpt-desktop.git (read-only, click to copy)
Package Base: chatgpt-desktop
Description: ChatGPT desktop application for Linux (repackaged from the official binary)
Upstream URL: https://chatgpt.com/download
Licenses: LicenseRef-OpenAI-Proprietary
Conflicts: chatgpt
Provides: chatgpt
Submitter: Akira-Amatsume
Maintainer: Akira-Amatsume (parsiad, mothran)
Last Packager: Akira-Amatsume
Votes: 40
Popularity: 16.65
First Submitted: 2026-08-15 13:05 (UTC)
Last Updated: 2026-10-02 17:11 (UTC)

Dependencies (43)

Required by (1)

Sources (4)

Latest Comments

1 2 3 4 5 6 .. 10 Next › Last »

potatoe commented on 2026-10-02 14:13 (UTC)

I don't know if this is old news but they are distributing Arch .tar.pkg.zst versions via a repository now, https://learn.chatgpt.com/docs/linux/linux-app#install-on-arch-linux .

Adding the keys and repository are done via an install script with so many belt-and-suspenders "don't break the user's setup" steps that it approaches obfuscation though.

Mr.Tao commented on 2026-09-26 13:13 (UTC) (edited on 2026-09-26 13:13 (UTC) by Mr.Tao)

I have an older version of this package installed from when it was called chatgpt-desktop-bin. I also preferred the previous UX for selecting models and reasoning to the current one. Also, the current package bundles everything instead of using the electron42 system package, making it twice the size of the original. I struggle to find out how the new package is better than the previous one.

Akira-Amatsume commented on 2026-09-18 19:00 (UTC) (edited on 2026-09-18 19:05 (UTC) by Akira-Amatsume)

@NonaSuomy Hyprland/Wayland is unrelated.

I originally set up Remote Control back when this package were still repackaged from the macos DMG. So I just revoked the connection and added it back, and it worked normally for me.

Given the `Couldn't check security requirements, try again' I think you are likely to be blocked by Cloudflare. There are some similar report:

https://github.com/openai/codex/issues/36450#issuecomment-5592679204

https://github.com/openai/codex/issues/31110

where /backend-api/wham/remote/control/mfa_requirement returned {"requirement":"required"} normally and /backend-api/accounts/mfa_info returned HTTP 403 with cf-mitigated: challenge and a Cloudflare managed-challenge HTML page. Since this request is made by a non-browser client, it can't complete the challenge, so the chatgpt app can't read the account's MFA state causing the client-side check for whether MFA is enabled to fail.

You can check it with:

curl -sS -H "Authorization: Bearer $(jq -r .tokens.access_token ~/.codex/auth.json)" \
  https://chatgpt.com/backend-api/wham/remote/control/mfa_requirement

grep -rn 'accounts/mfa_info' ~/.local/state/codex/logs/ | cut -c1-400

for more details.

While on my device, 'GET /backend-api/wham/remote/control/mfa_requirement' returns {"requirement":"suggested"}, thus won't check mfa_info, so I can't reproduce it for now.

Could you post some more logs ,or try a different network/public ip to see if the error still appears (btw are you using any proxy/vpn?).

NonaSuomy commented on 2026-09-18 13:26 (UTC) (edited on 2026-09-18 13:27 (UTC) by NonaSuomy)

Added MFA to the account still doesn't seem to work.

Have you personally tried it?

Settings -> Coding -> Connections -> Control This PC -> Add

"Couldn't check security requirements, try again"

I'm using hyprland + wayland if that has anything to do with it.

Assuming I'm missing something that that requires.

NonaSuomy commented on 2026-09-10 22:48 (UTC) (edited on 2026-09-10 22:49 (UTC) by NonaSuomy)

@Akira-Amatsume thank you, that was exactly the issue there was a stuck process. The UI crashed but the backend was still running. So even though chatgpt ui was showing the latest version the backend wasn't.

I just added MFA to the website but can't restart my client my local client to check just yet after that will check again.

You can probably unpin that message above as that old version being up there is confusing kept thinking I would have to downgrade to get astra working...

Akira-Amatsume commented on 2026-09-07 07:01 (UTC) (edited on 2026-09-07 07:16 (UTC) by Akira-Amatsume)

@NonaSuomy

  • chatgpt-desktop bundles its own codex-cli. For 26.901.51231, that's 0.153.4, so you don't need to install Codex separately. gpt-6-astra needs Codex version header 0.153.0 or newer, you can check the installed backend version and whether CODEX_CLI_PATH is set with:
/usr/lib/chatgpt/resources/codex --version
printenv CODEX_CLI_PATH

BTW, have you fully quit the app since updating it? Closing the window can still leave it running in the background, so it might still be using the old backend. Please quit it from the tray, reopen it and try again.

  • For Remote, that message means the app couldn't check your account's MFA requirements/status. You can check whether MFA is enabled on the chatgpt website under Settings → Security and login.

It could also be an upstream issue. There's a similar report where a Cloudflare challenge caused the MFA status check to fail with HTTP 403.

NonaSuomy commented on 2026-09-07 05:31 (UTC) (edited on 2026-09-07 05:42 (UTC) by NonaSuomy)

Also any ideas how to get the remote working?

Settings -> Coding -> Connections -> Control This PC -> Add

"Couldn't check security requirements, try again"

NonaSuomy commented on 2026-09-07 05:30 (UTC) (edited on 2026-09-07 05:32 (UTC) by NonaSuomy)

Seems it says I'm using 26.901.51231-1 but when I select Astra as the model it says I need a newer version... Any ideas why?

The 'gpt-6-astra' model requires a newer version of Codex. Please upgrade to the latest app or CLI and try again.

Is Codex a separate module?

Akira-Amatsume commented on 2026-09-05 16:21 (UTC)

Rechecked the upstream deb: Astra's visibility is set to hide in 26.901.41123. if you can't find it in your model list, upgrade to 26.901.41600 or run:

sed -i '0,/^\[/s/^model[[:space:]]*=.*/model = "gpt-6-astra"/' ~/.codex/config.toml

Thanks @Firestar for the feedback

Akira-Amatsume commented on 2026-09-05 06:54 (UTC)

@Firestar If you mean that astra doesn't show up at all, that depends on OpenAI's rollout schedule. As for me, now astra is already available in version 26.901.41123.

If you mean that astra isn't selected as the default model, that's determined by your ~/.codex/config.toml, not by this package. Also, once you've selected gpt-6 in a conversation, new conversations should automatically use it as well.

1 2 3 4 5 6 .. 10 Next › Last »