I tested the new Claude Desktop on Linux – here’s how it compares to rival apps


claudedesktophero

Jack Wallen/ZDNET

Follow ZDNET: Add us as a preferred source on Google.


ZDNET’s key takeaways

  • Claude Code finally has a Linux desktop app.
  • The app works swimmingly if you play by the rules.
  • Trying to make it work locally was an exercise in futility.

As a hard-core Linux user, I’m always on the lookout for apps that enable Linux to keep pace with competing platforms. You might be surprised that, even in this age of AI, competitive open-source options are not hard to find. For example, I’ve been using both Alpaca and Moose for some time without issue. Both of these apps serve as GUIs for instances of locally installed Ollama (which is how I typically like to roll with AI). They offer well-designed GUIs, the flexibility to work locally or in the cloud, and efficient use of system resources.

But sometimes, I yearn to work with a more mainstream option.

Such is the case with the newly released Claude Code Linux desktop app, which offers all the features found in the MacOS and Windows versions, and even lets you enable developer options to expand the feature set.

Also: I quit ChatGPT for a free, private, and local AI called Ollama – here’s why

Before I get into this, it’s important to understand that when I use AI, I almost always use it locally. I’ve installed Ollama on most of my machines I have. Why? I’m not straining the power grid or compromising my privacy. I will always choose local AI over cloud-based.

That said, I still wanted to kick the tires of the Claude Code Linux desktop app to see how it compared to the competition. 

Here are the important bits for me:

  • The app has to have a well-designed GUI that makes interacting with models simple.
  • The app needs to easily connect to locally installed AI (such as Ollama).
  • The app needs to function exactly as it does on other operating systems.
  • The app needs to use system resources wisely.

Does the Claude Code Linux desktop app live up to that checklist? Let’s find out.

Installing Claude Code on Linux

First things first: Install Claude Code on Linux. With many GUI AI apps, I can open the app store of my distribution, search for the app (such as Alpaca or Moose), and install. With Claude Code, I had to add the necessary repository before installation. One thing to keep in mind: Currently, Claude Code desktop is only available for Debian and Ubuntu-based distributions. 

Also: How to install and configure Claude Code, step by step

Here are the commands required:

Add Anthropic’s signing key: sudo curl -fsSLo /usr/share/keyrings/claude-desktop-archive-keyring.asc https://downloads.claude.ai/claude-desktop/key.asc

Add the repository: echo “deb [arch=amd64,arm64 signed-by=/usr/share/keyrings/claude-desktop-archive-keyring.asc] https://downloads.claude.ai/claude-desktop/apt/stable stable main” | sudo tee /etc/apt/sources.list.d/claude-desktop.list

Update and install: sudo apt update && sudo apt install claude-desktop

With the installation complete, I opened the Claude Code app from my desktop menu and was greeted — you guessed it — by a well-designed GUI.

Claude Code

The Claude Code Linux desktop app has a well-designed GUI.

Screenshot by Jack Wallen/ZDNET

It’s time to dive in.

Connecting to locally installed AI

This is where things got a bit bumpy (which wasn’t that surprising). I’ve set up plenty of local AI for various desktop clients, but (sadly) Claude Code was not the easiest. In fact, it was rather complicated, and I wouldn’t have been able to make it work without also installing Claude Code from within Ollama itself (with the ollama launch claude command, which allowed me to select an LLM to download and use with Claude Code).

I downloaded the Qwen6 LLM, which is 15 GB, so be careful what you wish for. Until you pull a model, the GUI won’t be able to find one.

Also: Want local vibe coding? This AI stack might replace Claude Code and Codex – for free

With that taken care of, I had to enable Developer options in Claude Code (Help > Troubleshooting > Enable Developer Options. After a relaunch, I could then go to Developer > Configure third-party inference. This is where things should have made it possible to connect to my locally installed AI. The settings look like this:

Claude Code

This should work as expected. Unfortunately…

Screenshot by Jack Wallen/ZDNET

Once you’ve done that, click Apply changes. You’ll then need to select a model for use. Scroll down to the Models section and click Add model. This is where I ran into what looked like an insurmountable hurdle. Even though I instructed Claude Code to use a local gateway base URL, it refused to see any of the models I’d pulled, and no matter how I tried to configure it, the Claude Code desktop app refused to comply. 

What this means is that I’m stuck with using a free Anthropic plan, which is pretty limited (especially when you try to get Claude Code to write an app for you). In the end, you cannot reliably use the Claude Code desktop app with local AI. Bummer.

Does the app function as on other operating systems?

This is where we get some good news. I tested the Linux app against the MacOS app, and they are identical. Feature for feature, you won’t find any difference (other than slight UI changes) between the two. 

Also: 5 reasons why I still prefer Perplexity over every other AI chatbot

Huzzah.

Does the app use system resources wisely?

This would have been a much more revealing test had the Claude Code desktop app functioned with locally installed AI. As is, it relied mostly on cloud-based resources, so using the desktop app didn’t put even the slightest dent in my machine.

Here’s the query I used for my testing:

Write a Linux GUI app for Pop!_OS COSMIC Desktop for creating invoices for clients. It needs to be able to keep track of clients and different deliverables (such as articles and videos), include custom fields, and export invoices to PDF documents.

Even though my free account couldn’t finish the query, while it worked, my machine continued performing fine. When I tested that against a locally installed AI desktop app that would function with local AI, the query brought my machine to a temporary halt while it ground away at the task.

Also: I’ve tested so many desktop AI tools, but Hermes with Ollama is my new favorite – here’s why

In the end, I was disappointed that I was unable to connect the Claude Code desktop app to my locally installed AI, but I can still say this: If you have an Anthropic account and are looking for a tool to simplify using Claude Code on your Linux desktop, this is definitely the way to go. I’ve launched Claude Code through Ollama (which does work quite well), but going that route is not for the faint of heart. 

If you want simplicity, this new Claude Code desktop app for Linux is a good option. Just make sure your Anthropic account is paid for and ready to go. If you want to use local AI on Linux, however, I’d suggest sticking to Alpaca or Moose.





Source link

Leave a Reply

Subscribe to Our Newsletter

Get our latest articles delivered straight to your inbox. No spam, we promise.

Recent Reviews







Today, when one pictures a “classic Dodge Charger”, the first image that pops up is almost certainly one of the highly desirable Charger models from the late 1960s or early ’70s. Indeed, those early muscle car Chargers are iconic, playing a starring role in the “Dukes of Hazzard” television show and, somewhat more recently, “The Fast and the Furious” films. But as time ticks on, is it time to start appreciating the modern version of the Charger as a potential modern classic?

It’s now been over 20 years since Dodge brought back the Charger nameplate for a spacious four-door sedan with an optional HEMI V8 engine. While the basic Charger R/T was a potent machine for its time, Dodge really took the Charger’s game to the next level for the 2006 model year with the debut of the Charger SRT8. 

The SRT8 model used a larger version of the third-gen HEMI V8 that, combined with other performance upgrades, transformed the sedan into a serious performance car capable of running with its 1960s HEMI ancestors at the drag strip — to say nothing of its vastly superior handling and refinement. In the years that followed, Dodge would continue to improve the Charger’s performance with larger and more powerful HEMI engines, but the significance of the original Charger SRT8 is not to be overlooked.

A muscle car legend reborn for the 2000s

Today, with the modern Charger being such an established part of the car enthusiast world, it’s easy to forget some of the controversy that surrounded its mid-2000s return. Most of it focused on the fact that the beloved muscle car nameplate had been brought back for a four-door sedan rather than a retro-styled coupe. Fortunately, those people looking for that retro coupe would be satisfied by the reborn Dodge Challenger when it arrived a few years later, while the Charger went on to become a highly popular muscle sedan in its own right.

The addition of the SRT8 model to the lineup certainly helped, of course. Under the hood was the larger 6.1-liter HEMI V8, which differed from the standard 5.7-liter HEMI in several ways, not least the displacement. With the 6.1 under the hood, the SRT8 made 425 hp and 420 lb-ft of torque, easily laying down a mid-13-second quarter-mile time in Motor Trend’s hands. This was very quick by mid-2000s standards, especially considering the now-outdated five-speed automatic transmission.

But the SRT8’s performance went beyond just the drag strip. As part of the SRT transformation, Dodge also gave the car larger wheels and tires, a retuned suspension setup, and large Brembo brakes. While this didn’t necessarily make the car an agile road course weapon, it did give the SRT8 an athleticism that belied the Charger’s weight and size. 

The evolution of modern Dodge muscle

What’s even cooler about this era in Chrysler/Dodge performance history is that the Charger was just one of the four-door LX platform cars that the automaker offered with SRT badges and a powerful HEMI engine under the hood. Apart from the Charger, buyers could also choose from the more upscale, but ultimately short-lived SRT version of the Chrysler 300C sedan or the Dodge Magnum SRT8 station wagon.

The original Charger SRT8 marked the beginning of a long run of increasingly powerful, high-performance models. In the early 2010s, the Charger SRT8’s 6.1 HEMI was replaced by the larger and more powerful 6.4/392 HEMI, with that motor eventually becoming available in the less expensive Charger R/T Scat Pack. Then, of course, came the Charger SRT Hellcat, with a 707-hp, supercharged 6.2-liter that turned the car into a genuine super sedan.

So is the original Charger SRT8 a guaranteed future classic? Classified listings show that clean examples still bring decent money today, but the fact that it was followed by improved models may ultimately limit its potential for becoming a true, mega-desirable collector car. Regardless, though, the Charger SRT8’s accomplishments in modern muscle car history are not to be taken lightly.





Source link