What your machine keeps.
A fixed, published price per unit of work. All of it to the machine that did the work. Change any assumption below; the arithmetic is the only thing here that is not an assumption.
PLANNED planned reference prices; illustrative throughput; the network does not run yet
$63 a month, to you
629 million tokens a month at $0.10 per million. Rutli keeps $0 of it.
The same work on a marketplace that keeps 25%: $47. At 30%: $44.
Illustrative. Throughput and the share of busy hours are assumptions you can change; the price is a planned reference price, set with the founding nodes. Your electricity, taxes, and pass-through payout costs are yours. Not a promise.
How it works, in three lines.
You set the rules.
Hours, a ceiling on the GPU, a power cap, the kinds of work you accept, the countries it may come from. Your own use comes first, always.
The network sends work to your rules.
A job arrives signed, runs in a sandbox you sized, and is interrupted the moment you need the machine. Metered per unit of work at the published price.
You receive 100% of the price.
On your balance, in ordinary money, with every line readable. Rutli keeps none of the compute price. Pass-through costs shown as themselves.
Questions
How does Rutli pay node owners?
A customer pays the published price for a unit of work. The machine that did the work receives 100% of it, credited to the owner's balance as the work completes, in ordinary money. Payment and payout costs are pass-through, at cost, shown as their own lines.
Who sets the price?
Rutli publishes a fixed price per unit of work for each model. Not a marketplace where hosts bid each other down: one price, public, changed only with a published date. Why: a stable, known price is what makes the decision to share a simple one, and it keeps the network from a race to the bottom that only the largest hosts survive.
How is utilisation handled?
The network schedules work to the hours, ceiling and power limit you set, and it does not promise to fill them. How much work reaches your machine depends on demand and on the machines available; the calculator makes you choose that share yourself, because nobody can promise it. The status page will publish what is measured.
Does my own use come first?
Always. A task of yours preempts network work at once; the job is interrupted and resumed elsewhere. Sit down to play, start a render, run your own inference: the network gets out of the way.
Homes and offices, or data centres?
Machines people already own, in homes, studios and offices. Windows, Linux and macOS, including Apple silicon. Nothing here needs a rack, a static address, or an open port on your router.
How quickly will my machine get work?
Honestly: the network does not run yet. The first machines to do work are the founding nodes, in the alpha, and the first outside workloads arrive with the first outside customers. This page shows the terms, not a queue.
Is there a fee on anything else?
No fee on the compute price, ever. Services Rutli itself sells, around the network, are priced separately and say so. Nothing is deducted from what your machine did.
Can I set my own price?
No. You set the hours, the ceiling, the power, the kinds of work you accept, and the countries the work may come from. The price is the network's, published, and the same for every machine of the same class.