I tried the new ChatGPT Desktop App for Linux – but I’ll stick to my browser for now

Must Read
bicycledays
bicycledayshttp://trendster.net
Please note: Most, if not all, of the articles published at this website were completed by Chat GPT (chat.openai.com) and/or copied and possibly remixed from other websites or Feedzy or WPeMatico or RSS Aggregrator or WP RSS Aggregrator. No copyright infringement is intended. If there are any copyright issues, please contact: bicycledays@yahoo.com.

Observe ZDNET: Add us as a most well-liked supply on Google.


ZDNET’s key takeaways

  • OpenAI’s ChatGPT Desktop App for Linux is an Electron-based program.
  • The applying is proprietary, not open supply.
  • Frankly, it is not that spectacular. I will maintain utilizing ChatGPT within the browser.

OpenAI has launched a preview of its ChatGPT desktop utility for Linux. With this, OpenAI now has a local consumer on all main desktop working programs and is inserting its coding agent, Codex, alongside ChatGPT and its Work surroundings.

That is the excellent news. The unhealthy information is that, for me at the very least, ChatGPT continues to be annoyingly inaccurate. Whereas I can cosplay a coder, with or with out AI, my actual use for AI is as a sophisticated search engine. After I requested the ChatGPT app on my fundamental Linux Mint desktop to offer me the technical specs for SpaceX and Nvidia’s forthcoming Starmind AI datacenter satellites — and offered it with a hyperlink to the satellite tv for pc’s website — I anticipated an correct reply. I did not get one. Starmind’s peak and wing-width, in response to SpaceX, are 30m and 75m, respectively; not, as ChatGPT would have it, 20m and 70m.

(Disclosure: Ziff Davis, ZDNET’s mum or dad firm, filed an April 2025 lawsuit in opposition to OpenAI, alleging it infringed Ziff Davis copyrights in coaching and working its AI programs.)

Supported Linux distros 

Leaving apart ChatGPT’s lack of accuracy, this launch helps Ubuntu 24.04 LTS and 26.04 LTS, Debian 13, and Fedora 43 and 44 on each x86-64 and Arm64 machines. OpenAI is distributing the consumer as DEB and RPM packages. In follow, this implies you possibly can run on most Debian and Crimson Hat-family Linux distributions, resembling my very own Linux Mint. 

The discharge additionally arrives roughly a month after Anthropic launched a Linux desktop model of Claude. Anthropic’s consumer helps Ubuntu 22.04 or later and Debian 12 or later, giving it a considerably broader said baseline amongst Debian-derived programs. 

Arch Linux, NixOS, and different distros are left within the lurch for now. That is made some would-be customers very sad. Others need this system to be distributed in Flatpak or AppImage codecs. 

Nonetheless, in response to OpenAI, since “Linux has been one of many most-requested platforms for the desktop app, and this launch extends ChatGPT and Codex throughout each main desktop working system,” I think about we’ll see different distros supported quickly. 

A developer-oriented desktop consumer

The Linux utility consolidates three OpenAI merchandise which have more and more overlapped. These are ChatGPT for general-purpose AI help, ChatGPT Work for delegated office duties, and Codex for software program growth. In accordance with OpenAI’s Linux preview announcement, the desktop app is meant to function a workspace for initiatives, information, browser workflows, and Codex periods. For Linux builders, Codex is the headline function.

That positioning issues in Linux, the place ChatGPT has been broadly accessible in a browser however the place builders have relied on operating purposes within the shell, API integrations, unofficial Electron wrappers, and editor extensions for extra native workflows. The brand new consumer may give OpenAI a extra direct foothold on developer desktops, as AI coding instruments turn out to be extra tightly built-in with native initiatives and growth environments.

For programmers, the headline options begin with the power to code native initiatives: Codex can study and modify information, run instructions, and work inside chosen mission directories, with conduct constrained by your chosen permission modes. For a neighborhood mission with a number of folders, the desktop app can present repositories and altered strains collectively for consolidated overview.

The broader desktop app function set additionally contains inline diff modifying and pull-request overview. Characteristic availability might fluctuate within the Linux preview, so OpenAI has not printed an in depth Linux-versus-MacOS/Home windows parity desk. For the fullest-featured model of the applying, you continue to want to make use of a Mac.

Lastly, whereas the app can use browser workflows, desktop integrations the place supported, and plugins, every linked service’s entry request should be reviewed and authorized individually.

As well as, regardless of operating on Linux, the consumer itself is proprietary and closed-source. That may make it a non-starter for a lot of open-source builders. 

At launch, the Linux consumer doesn’t, by default:

  • Routinely scan all native disks, folders, or purposes.
  • Management arbitrary Linux GUI purposes by native pc use.
  • Entry an internet service merely as a result of you’ve gotten an account there. You have to give it a certified browser connection, plugin/integration, uploaded information, or an specific activity/motion.
  • Acquire root or administrator authority simply because it might probably run instructions in a particular coding workspace.

Integration wins

I am not too trusting of letting any AI program entry my native and private packages. Nonetheless, if you happen to’re OK with that, go forward and combine ChatGPT Desktop App with Google Calendar, Gmail, Outlook, Groups, Slack, and Google Drive. As soon as that is finished, you possibly can then have it examine your linked messages, e-mail, calendar, paperwork, and mission trackers. You may as well pull information and duties out of your Claude workflow.

Armed with that data, it might probably monitor modifications, put together updates, and draft replies. You may as well use it to create or coordinate a separate activity for a selected mission. 

Take into account, nevertheless, if you happen to do this, ChatGPT will get entry to all of your uncovered information. Particularly, something you sort, paste, add, or intentionally add as context is shipped to ChatGPT for processing. By default on shopper accounts, conversations will  be used to enhance OpenAI’s fashions until you flip off Settings > Knowledge Controls > “Enhance the mannequin for everybody.” That setting applies account-wide. Chats stay in your historical past however aren’t used for mannequin coaching, OpenAI says.

All that’s cause sufficient for me to keep away from this type of performance regardless of how helpful it might sound.

Oh, and for now anyway, OpenAI says native “pc use” — controlling different Linux desktop purposes — is at present unavailable on this preview. That is tremendous by me. It does, nevertheless, offer you entry to a wide range of different companies by way of plugins, resembling Figma, GitHub, and Adobe. 

Wise Linux setup

If you are going to use the desktop app to your private or necessary information, I like to recommend taking the next steps to make use of it as safely as potential. 

  • Hold “Enhance the mannequin for everybody” off, until you need chats used for mannequin enchancment.
  • Use short-term chats for particularly delicate one-off queries, recognizing the said 30-day deletion window.
  • Give it a devoted, slim mission listing. Do NOT give it entry to  ~/, your whole paperwork tree, or a listing holding credentials and supply information.
  • Depart Codex on Ask for approval. Don’t allow Full entry until a activity actually requires it.
  • Use a separate browser profile with solely the service accounts required for an agent activity. Keep away from operating it whereas password managers, personal e-mail, banking, or confidential supply materials are open.

I would add that it is best to take these steps no matter what working system you are utilizing. Certain, it is simpler to make use of AI with out taking precautions, nevertheless it’s not secure. 

Hassle in Linux ChatGPT land

This new app nonetheless has an extended technique to go. For instance, I discovered that my ChatGPT shell jobs would not present up on the desktop app’s mission checklist. That is a fairly main stopper proper there. It wasn’t simply me. I noticed others reporting the identical downside. You would see terminal periods as standalone chats, however even when the working listing and Git department had been preserved, they actually weren’t built-in. 

I additionally found that even when sitting idle with no work to do, the ChatGPT app used over a gigabyte of RAM. Including insult to damage, I found this system leaked reminiscence. After only some hours of very mild work, my PC, with 16 GB of RAM, was operating out of reminiscence. As soon as my pc began calling on digital reminiscence, it slowed to an unusable crawl. 

That is not too stunning since. After a bit digging, I discovered it was an Electron app. These are desktop software program packages written with net applied sciences like HTML, CSS, and JavaScript. They mix the Chromium browser engine for the person interface and Node.js for system entry, permitting a single codebase to run on Home windows, MacOS, and Linux. That is tremendous so far as it goes, however Electron packages are usually terribly gradual and memory-inefficient in comparison with native working system purposes. 

Early Wayland points emerge

I have never seen it myself, however preliminary group suggestions factors to the acquainted challenges of delivery a desktop utility throughout Linux’s assorted show servers.

The Wayland window supervisor is a specific ache level. One Fedora 44 KDE Plasma person reported that Japanese enter utilizing Fcitx 5 didn’t work within the ChatGPT message composer when launching the applying usually. The person stated the difficulty might be addressed by beginning the app with Chromium/Wayland flags, together with `–ozone-platform=wayland` and `–enable-wayland-ime`. This permits the consumer to stay on Wayland slightly than falling again to XWayland. Different discussion board members reported comparable conduct with Korean Fcitx5 enter and raised considerations about X11 fallback and scaling on high-resolution Wayland desktops. 

You’ll be able to’t say ChatGPT did not warn you. The applying’s web page now says, “Native Wayland help is experimental and can proceed to enhance. In a Wayland session, the app makes use of XWayland when accessible. To explicitly choose native Wayland, totally give up the app and launch it from a terminal.”

All this underscores the distinction between supporting a number of distributions on paper and integrating cleanly with Linux’s big selection of compositors, enter strategies, and workflows.

Shared issues

Considering of workflows, my ZDNET colleague David Gewritz lately identified that the newest model of ChatGPT Desktop on the Mac is now lacking two options he discovered very helpful. These are the power to take screenshots from inside the app and instantly drop them into chat, and “Work with,” which helps you to feed ChatGPT the contents of your present window from apps resembling Notion, Notes, and TextEdit. They are not within the Linux model both.

Behind the scenes, Gewritz additionally discovered that you just’re usually channeled into higher-token-priced service ranges. You do not have to pay for these, however if you happen to do not, your efficiency will decelerate. The day of low cost AI is finished. 

Strategic significance

The Linux consumer provides OpenAI a local path to the Linux builders who create the majority of AI infrastructure and AI packages. The app’s central function shouldn’t be merely a ChatGPT window indifferent from a browser. It is the combination of Codex and work-oriented agent options right into a desktop workspace.

Actually, I have never been impressed. I am positive it is going to get higher, though for an organization with OpenAI’s billions, I am shocked they got here up with such a half-baked Electron utility. Absolutely, they might do higher. Within the meantime, after I use ChatGPT, I will maintain utilizing it with my browser. 

Latest Articles

AI Cites the Same Papers Over and Over Again – Just...

ChatGPT and different AI writing instruments could also be quietly turning science right into a reputation contest, repeatedly steering...

More Articles Like This