Lend your graphics card to the swarm and, while it's on, use the community's models at no cost, including the ones your card can't load on its own. Every card that joins brings the next level closer, all the way to a 70B model. No credit card needed.
What is running right now on the lent cards, and what each level still needs. Not a promise: this is the real data, live.
Checking the swarm…
A split model needs cards close to each other (adding up memory isn't enough), and over the internet it runs slower than on a single card. Updates every 30 seconds.
A Windows app that lives in the system tray. Install Ollama, pull a model and the client detects it automatically.
You start by lending half. Raise it, lower it or switch it off whenever you want to game or work: your PC is still yours.
Once your card is approved, your requests to the swarm cost you nothing while it's on, including the models your card couldn't load on its own. You can see it on your dashboard.
Together with other nearby cards, the swarm loads models that don't fit on one, up to 70B.
By default you lend half of your card; you decide whether to give more or less.
Turn it on, off or limit the power whenever you want. No contracts, no lock-in.
Your machine only runs AI jobs signed by the operator; the client gives no access to your files.
Using the swarm's models, the ones running on the community's cards, while your card is lent; if you switch it off, you lose it. Cloud models are paid separately. Which models are available depends on how many cards are on (you can see it live above), and a model split over the internet runs slower than one on a single card. The more of us, the better for everyone.
Prefer to share your GPU only with people you trust? Create a private group: only your friends use it and everyone pays their own way.
Create a private groupCreating an account is free. Lend whenever you want and stop whenever you want.
Join the community