Skip to main content
Most of what Wally runs is hosted — see the Cloud quickstart. When a small model on your own machine is enough, the same CLI runs it, offline and without an account.

​
1. Install

Check it landed:

​
2. Pull a model

About 640 MB. wally models list --all shows everything available, and wally models list shows what you already have.

​
3. Run it

That opens an interactive session. To ask one question and exit:

​
Where to go next

Integrations

Point opencode at the model you just pulled.

Wally Cloud

Frontier-scale models we host, through the same commands.

On-device SDK

Ship this inside your own app, on six platforms.

Serve it over HTTP

wally serve qwen3-0.6b serves the model you just pulled on a local OpenAI-compatible URL.