Get started#
Astralyx turns the machines you have into one AI cloud: a laptop without a GPU, a gaming PC, a Mac, a home server, racks in a datacenter. Pick what you have. Each path grows into the next without changing tools — what you learn on one machine is how the datacenter works too.
-
One machine start here
A Linux machine or a Windows PC (through WSL 2) — with or without a GPU. Create an account, add the machine with one command, and run a model on it in about ten minutes. Macs, and one-click apps for Mac and Windows: coming soon
-
A few machines home lab, office
Run the same command on each: they join the same cluster and their GPUs add up. A model runs on the machine it fits; the rest keep working.
-
Your own servers production
Run Astralyx's services on your servers: signed packages for any Linux, one command with a key from the console, nothing that connects in.
-
Everything at home coming soon
One app or one command, and your machine keeps a private cluster of its own: nothing you run reaches the internet; you use it from your network.
-
Not sure?
How each way of installing works, what it gives you and what it asks of you, side by side.
What one machine can do#
You do not need a GPU to start.
| The machine | What it runs well |
|---|---|
| A laptop or server without a GPU | Small models on the CPU (up to ~3B parameters), assistants, functions, workflows, knowledge bases |
| A Mac with Apple Silicon (coming soon) | Models on its GPU, sized by its memory (a 32 GB Mac runs models up to ~30B) |
| A PC with an NVIDIA or AMD GPU | Models sized by its GPU memory (12 GB: up to ~14B), image generation, fine-tuning small models |
| Several machines | One model per machine, each where it fits, all behind one API |
The console's model catalog says, for each model, on which of your machines it fits — and why not, when it does not.