I report my first experiences with LM Studio in (Arch) Linux.

LM Studio is a desktop application that allows you to discover, download, and run Large Language Models (LLMs) entirely on your own computers, offline.

This is the hardware and system I’m testing it (you see, no GPU with VRAM):

First, download the AppImage from the website https://lmstudio.ai/.

Run it, and follow the welcome steps:

It proposes to download the first model, which seems to fit my hardware (Gemma 4 e4b):

Finally, we get to the main interface:

Now, select the model (the only one currently downloaded)

And you can start to chat (in the following, I’m experimenting with creating a Maven project, but I’ll blog about such experiments in the future):

You can configure model parameters and possibly make such a configuration the default for the model:

For example, increase the context length:

Download other models using the UI for searching:

Note how LM Studio tells whether a model fits your hardware or is likely too large:

Alternatively, use the CLI “lms” to download a model:

During the downloading, the GUI also reports that:

Note that models are downloaded by default into “~/.lmstudio/models”:

OK, that’s all for this post!

Stay tuned for experiments with LM Studio and open source models.