Axom is almost here.
Public downloads open soon. When they do, it's one command on macOS or Linux, and Axom sets itself up for the machine you have.
One command. It handles the rest.
Install
One line in your terminal. Axom installs into its own private environment — no admin rights, nothing else on your system touched.
Scan
Axom reads your GPU and memory, shows you every model tier, and recommends the one that fits.
Verify
Every file is checked against its published SHA-256 before it runs. A mismatch is refused, not installed.
Stay current
axom update finds the newest release, verifies it, installs it, and reloads — your memory and settings stay put.
One architecture. Sized to your machine.
Every tier runs the same Axom — the same memory, tools, and leaves — on a Gemma 4 chassis, American-made and unmodified.
| Tier | Chassis | Built for | GPU | or RAM |
|---|---|---|---|---|
| X1-Ultra | 26B mixture · 4B active | The flagship. Big-model quality at small-model speed. | 20 GB | 32 GB |
| X1-Max | 12B | Serious workstations. | 16 GB | 24 GB |
| X1-Heavy | 4B · full precision | Full quality on everyday hardware. | 10 GB | 16 GB |
| X1-Light | 4B · compact | The same model at half the size. | 7 GB | 10 GB |
| X1-Nano | 2B | Fits almost anywhere. | 5 GB | 8 GB |
Apple Silicon counts its unified memory. No GPU? Axom still runs on CPU — or on the Axom Network, where a node does the heavy lifting and your memory and files stay on your machine.