You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I used os versioning on this as lm studio lacks any kind of api to get versioning I could find.
This has no aarch64 variant as the image is 8 gigs, down the line I might revisit this as it can be aarch64 but I think there is minimal demand for that as of now given the state of arm GPU support.
This is an Arch KDE desktop with LM Studio and llmster setup to be launched from the desktop, the user can just use the desktop app as is or use it to manage their local models. If they really know what they are doing they can use lms from the CLI to ingest models as well. From there you can launch llmster which is the headless server and start using the agentic tools like opencode, cline, or openclaw that are all baked in. All agentic tools have been fully tested on AMD and Nvidia. It works on Intel but the Vulkan performance at least for me was basically the same as CPU so I am not advertising that.
ROCm is in Arch repos but is 27 gigs so I left that out as Vulkan works great for me. It looks like ROCm is more used for diffusion support now. This only supports Cuda v13 and my recommendation for Nvidia users has always been to install the latest binary drivers host level.
Tested 3060 12gb and Radeon RX-6700 on Debian Trixie install.
Tested my AMD BC-250 setup as well.
Opencode on my tower:
2026-06-23.10-54-33.mp4
Having the full desktop like this not only allows people to run LM Studio but it also enables more complex agentic workflows in an actual sandbox where it does not matter what it messes up, even during my testing OpenClaw managed to kill a bunch of random pids and delete files it was not supposed to.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
I used os versioning on this as lm studio lacks any kind of api to get versioning I could find.
This has no aarch64 variant as the image is 8 gigs, down the line I might revisit this as it can be aarch64 but I think there is minimal demand for that as of now given the state of arm GPU support.
This is an Arch KDE desktop with LM Studio and llmster setup to be launched from the desktop, the user can just use the desktop app as is or use it to manage their local models. If they really know what they are doing they can use lms from the CLI to ingest models as well. From there you can launch llmster which is the headless server and start using the agentic tools like opencode, cline, or openclaw that are all baked in. All agentic tools have been fully tested on AMD and Nvidia. It works on Intel but the Vulkan performance at least for me was basically the same as CPU so I am not advertising that.
ROCm is in Arch repos but is 27 gigs so I left that out as Vulkan works great for me. It looks like ROCm is more used for diffusion support now. This only supports Cuda v13 and my recommendation for Nvidia users has always been to install the latest binary drivers host level.
Tested 3060 12gb and Radeon RX-6700 on Debian Trixie install.
Tested my AMD BC-250 setup as well.
Opencode on my tower:
2026-06-23.10-54-33.mp4
Having the full desktop like this not only allows people to run LM Studio but it also enables more complex agentic workflows in an actual sandbox where it does not matter what it messes up, even during my testing OpenClaw managed to kill a bunch of random pids and delete files it was not supposed to.