Cascadia local

Run local AI models on the Intel PC you already own.

Run local models fully offline on the Intel hardware you own, from one machine to several. No cloud, no subscription, nothing leaving your network.

live in open alpha today

Why run local models?

Private by default

Your prompts and documents never leave your devices.

Yours to own

Runs on hardware you bought. No account required.

No subscription

No per-token bill. Generate as much as you like.

Works offline

On a plane, in a SCIF, off the grid. No signal needed.

A person working on a laptop before a vast cosmic cave, alone and off-grid.

An honest look at one machine.

7B-8B (Q4)
on a Core Ultra iGPU~17-18 tok/s
good forchat, drafting, summarizing
up to 14B
on a Core Ultra iGPU~28-35 tok/s class
good forcoding help, longer reasoning
500+ GiB / 1T
on a Core Ultra iGPUpool several machines
good forfrontier reasoning, on-prem

Three steps to run local models.

01

Install

One app on your Intel Core Ultra PCs.

02

Pick a model

Any open Hugging Face transformer, optimized for Intel.

03

Run offline

Disconnect. It keeps working.

running a team?

Local already pools the Intel machines on your network. Cascadia Enterprise adds what a team needs on top: user management, security policies, and central control, up to a 1T model, on-prem.

The same engine, with enterprise controls.

Explore Enterprise

Get capable AI on your own machine.

Live in open alpha today. Download it and run capable local models on the Intel hardware you already own.

Free for individuals

Coming soon

Deploying across an organization?

Talk to our team