Bring Spike Local home.
Download the Windows preview today. macOS and Linux builds are on the way — join the waitlist and we'll send your link the moment they land.
Get on the list.
One email when your build is ready — plus the macOS and Linux versions as they ship. No spam, no account, unsubscribe anytime.
What Spike Local needs to run.
Spike runs a real local model on your machine, so the experience scales with your hardware. Here's the honest picture.
Technically works.
- Windows 10/11 (64-bit) with WSL2
- 8 GB RAM
- ~10 GB free disk for the engine + a starter model
- Runs CPU-only — no GPU required
Actually impressive.
- 16 GB+ RAM
- A 7B-class local model for sharper output
- A capable GPU — or a GPU workstation on your LAN via the remote model host
- Apple Silicon for a strong fully-local experience (soon)
The gap between "technically works" and "actually impressive" is mostly model size. Spike runs on modest hardware, but a larger local model — or a stronger machine on your network doing the heavy lifting — is what makes it feel magical.
From download to first answer.
Install Spike Local
Run the installer. On Windows, Spike sets up its local engine in WSL2 — guided, with a System Check screen that flags anything missing.
Let it provision
The first run pulls the local model and speech tools (a few minutes). One-click fixes handle any missing components.
Point it at your world
Pick a folder or answer a few questions. Spike indexes your files and builds its first picture of your work.
Ask something real
A draft in your voice, an answer from your files, minutes from your meeting — the moment Spike becomes yours.
Your assistant is waiting on your machine.
Download the Windows preview, or join the waitlist for the macOS and Linux builds.