For NVIDIA graphics cards with 16 GB or more
An NVIDIA card with 16 GB of memory or more can serve the open models developers and their agents call through our API. Hosting is opening in stages: join the list, and we invite cards in the order paid demand fits them, after a short review.
Small and mid-size chat, vision and embedding models: the fast, cheap tier that answers most requests.
The larger chat and agent models, with tool calling and long context: what agents ask for most.
The largest models we serve, with room for long context and several requests at once.
Joining needs a free provider account (one click with Discord, Google, GitHub or Twitch). Then join from your dashboard, or come back to this page.
Soon, if it is an NVIDIA card with 16 GB of memory or more. Model hosting opens to provider cards in stages, after a short review, as paid demand for each card size arrives. Join the list on this page and we invite cards in that order.
When paying customers need more of a model than our own machines serve. We don't promise a date: we would rather invite your card when there is work for it than have it sit idle.
A share of every token billed on your card, set by your payout tier: the same tiers as every other job on the network, shown on your dashboard. Earnings are checked, held for the usual clearing period, then yours as credit or cash.
No. Joining the list needs a free provider account and a minute. If you want your card earning meanwhile, install legit2node and it takes clip, caption, transcription, music and sound-effect jobs while you are away.
Requests from customers who switched on Private never go to provider cards; they stay on our own machines. Hosted models serve the shared tier, and prompts are never stored.
NVIDIA cards with 16 GB or more on Windows or Linux. The larger the card, the larger the models it can serve; 24 GB and up covers the models agents ask for most. AMD and Intel graphics are not supported yet.
Earning with your card today: what it takes and how you are paid.