Installation

Coming soon

Public downloads aren't open yet. This page describes what installation will look like. Follow @axom_labs for launch.

Requirements

• macOS or Linux
• Python 3.10 or newer
• A GPU is optional. Axom picks the model tier that fits your machine, and falls back to CPU when there's no GPU.

The X1 lineup

Every tier runs the same Axom, with the same memory, tools, and leaves, on a Gemma 4 chassis.

Tier	Chassis	GPU memory	or RAM
X1-Ultra	26B mixture of experts · ~4B active	20 GB	32 GB
X1-Max	12B	16 GB	24 GB
X1-Heavy	4B · Q8	10 GB	16 GB
X1-Light	4B · Q4	7 GB	10 GB
X1-Nano	2B	5 GB	8 GB

On Apple Silicon, Axom counts unified memory. Below every tier's minimum, setup offers the Axom Network instead.

What the installer does

1. Creates a private Python environment in ~/.axom/venv and puts the axom command in ~/.local/bin. No admin rights needed; nothing else on your system is touched.
2. Verifies the package against its published SHA-256 checksum before installing it. A mismatch is refused.
3. On Linux without an NVIDIA GPU, installs the small CPU build of PyTorch instead of the multi-gigabyte CUDA build.
4. If Axom is already running, reloads it on the new version.

Updating

axom update          # install the newest release, verified, then reload
axom update --check  # only report whether a newer version exists

Your memory, settings, and history stay where they are.