Turn an idle GPU into income
Run the QueraIS node, serve open-weight LLMs to the marketplace, and keep 95% of every job — paid in $QAIS. You set your own price; the network routes work to you based on price, reputation, and speed. Like BitTorrent seeding, but you get paid.
Install in one line
One command installs Node and Ollama if you need them, downloads and checksum-verifies the latest node, sets up a working config, and starts serving — nothing to edit, no second step. Nodes join the live Arbitrum One network; your wallet stays on your machine.
iwr -useb https://inference.querais.xyz/install.ps1 | iexcurl -fsSL https://inference.querais.xyz/install.sh | shPrefer to do it by hand? Grab the archive from GitHub Releases and run the launcher — same result.
Requirements
- GPU with 8 GB+ VRAM to serve a tier
- Node.js + Ollama — installed automatically if missing
- $QAIS stake for your tier (from 1,000) — acquired by direct arrangement, no faucet
- A little ETH on Arbitrum One for registration and claim gas
- Wallet generated locally on your machine — keys never leave it
Estimate your earnings
Hardware tiers
From a gaming card to a datacenter GPU — your hardware decides which models you can serve and how many jobs you win.
| Tier | GPU | VRAM | Serves | Est. earnings |
|---|---|---|---|---|
| Bronze | GTX 1070 / RX 5700 | 8 GB | Llama-3-8B, Mistral-7B (Q4) | $7–25 / mo |
| Silver | RTX 3080 / 4070 | 12–16 GB | 8B FP16, 13B (Q8), Qwen2-7B | $25–95 / mo |
| Gold | RTX 4090 / 2× 3090 | 24 GB+ | All Silver + Llama-3-70B (Q4) | $75–255 / mo |
| Platinum | A100 / A6000 / H100 | 40 GB+ | Everything, full precision | $400–4,000+ / mo |
Earnings are rough estimates at 30–40% utilization and current network rates — not a promise. Real income depends on demand, your price, uptime, and reputation.
Live in about ten minutes
Install. Run the one-liner above. It installs Node + Ollama if missing, downloads the node, generates an encrypted wallet, and writes a working config — no editing.
Connect a wallet. Create a fresh wallet or import one — it receives your $QAIS earnings.
Stake. Lock the minimum $QAIS for your tier (from 1,000). Acquire $QAIS by direct arrangement with the operator — there is no faucet — and keep some ETH for Arbitrum One gas. Stake is your collateral and unlocks higher-value jobs.
Pick models. Choose which open models to serve; downloads resume in the background and are SHA-256 verified.
Go live. Your node announces over libp2p and starts bidding — appearing in the marketplace within a couple of minutes.
🎉 Your node is LIVE
status: Active
tier: Silver (stake 5,000 QAIS)
models: llama3.2, mistral-7b
dashboard: http://localhost:3000
today: $4.28 · this month: $127.50How you earn
95% to you
Each settled job pays the serving node 95%; the protocol keeps 5%.
You set the price
Quote your own per-token price, or let the optional auto-pricer track market and load.
Batched payouts
Earnings accrue in escrow and settle in batches — claim anytime for cents of gas.
What the daemon runs
A single binary handles everything — model management, inference, bidding, P2P, payments, and a local dashboard at localhost:3000.
Inference
llama.cpp by default (GGUF), with optional vLLM / ExLlamaV2 for advanced setups.
Model manager
Registry sync, resumable downloads, SHA-256 integrity, dynamic VRAM load/unload.
Job handler
WebSocket job feed, auto-bid calculator, result packaging, completion reporting.
Safe by default
Auto-pauses bidding when your GPU is busy with other work; staking risk is bounded by clear slashing rules.