The cheap, open alternative for AI: chat, image/video/audio generation, and a developer API — all served by a worldwide network of community GPUs. Pay with crypto. Earn by plugging yours in.
// Windows may warn "unknown publisher" on new apps, click Keep, then Run anyway
You don't come here to pay Big Tech prices. Open models — chat, media and an API — served cheap by a community GPU network. Enough for most work, at a fraction of the cost.
Multi-model chat plus n1o Code, an agent that works in your real files. Open models, far cheaper than the closed labs.
Image, video, audio and voice from open models — the kind that's gated or pricey elsewhere. The GPU network is built for it.
One endpoint for open models (Llama/Qwen-class): cheap, usage-based, no lock-in. Build your app on inference you can actually afford.
Wallet login, pay with crypto, no data harvesting. Community-owned — and you can earn by contributing your GPU.
What BitTorrent did for files, n1o does for intelligence. Large models live in billion-dollar data centers. We split them instead: each person hosts a slice of the model on their own GPU, and answers travel across the swarm. Nobody owns it whole, everybody runs it.
One click in the desktop app. A slider sets how much memory you contribute, the rest stays yours for gaming and work.
Every request to the community model travels through the network's GPUs. Open models, owned by everyone, owned by no one.
Proof of Inference: n1o points for every token YOUR card actually processes, not for sitting idle. A future $n1o token for redemption is being explored — it may never launch.
// experimental — points are an internal record of useful work; any future $n1o token or redemption is exploratory, not a promise, and may never happen
The n1o network aggregates people-powered GPU capacity into a single programmable layer. AI companies use it as complementary infrastructure, alongside their own data centers, not instead of them.
Native builds for the three major platforms, signed and self-updating. Install once — every new version arrives on its own.
The n1o headless node runs on any rig with a terminal: no screen, no reinstall, no new OS. Two commands and every GPU in the farm earns for the AI work it serves.
Your worker → Remote Access → Hive Shell. Paste the commands below and the node starts with the rig from then on (systemd).
Workers → your rig → Console. Same commands — python3 is already there.
Ubuntu, Debian, whatever your farm runs. If you can SSH in, you are 30 seconds away.
// disk: to just contribute VRAM the agent is under 1 MB, any disk works. To serve the model (and earn work points) the rig downloads it once: ~5 GB for an 8-GB card, ~9 GB for 12 GB, ~20 GB for 24 GB. A 32 GB system disk covers every tier up to Pro.
// YOUR_TOKEN lives in app.n1o.ai → n1o Network → "Vincular por consola" · points go to your account, one token links your whole farm
// reparto: cada GPU (--rpc) NO guarda el modelo, solo ejecuta sus capas en VRAM. La red elige sola qué nodo orquesta (director) — nadie lo teclea — y ese baja el modelo y lo trocea entre todas. Un modelo de 100 GB entre 15 GPUs de 8 GB = nadie tiene el modelo entero. Cada GPU cobra lo suyo. Funciona entre casas distintas por el relay del VPS (aunque haya router/NAT).