Shipping the testnet
Update 2026-08-23: QueraIS is now live on Arbitrum One mainnet — the gateway, contracts, and on-chain settlement described below run in production today. There is no faucet; $QAIS is acquired via direct arrangement only. This post is preserved as written on its date.
The QueraIS testnet is live on Arbitrum Sepolia. Here's what actually works today.
Call it like OpenAI
Change one line — the base URL — and your existing OpenAI code runs on the network:
from openai import OpenAI
client = OpenAI(base_url="https://gateway.querais.xyz/v1", api_key="sk-...")
r = client.chat.completions.create(
model="llama3.2",
messages=[{"role": "user", "content": "Explain Arbitrum in one sentence."}],
)
print(r.choices[0].message.content)
Streaming works, a model picker works, and there are official SDKs — @querais/sdk on npm
and querais on PyPI — that add the marketplace surface (routing options, node/stats
introspection, batched-settlement sessions) on top of the OpenAI protocol.
Run a node in one command
Got a GPU and Ollama? One launcher installs Node and Ollama if you need them, downloads and checksum-verifies the node, auto-claims $QAIS from the faucet, stakes, registers on-chain, and starts serving — no config to edit. You keep 95% of every job.
Settlement that doesn't nickel-and-dime you
Pre-fund a credit account, sign one EIP-712 spending cap, then fire thousands of calls — they batch-settle in a single on-chain transaction, so per-call gas is effectively zero. The gateway can only ever settle at the prices in your signed job specs, never beyond your cap.
What's next
The trusted-gateway model is the starting point, not the destination — on-chain auction, a P2P mesh, and a decentralized oracle remove it piece by piece. Follow along on the roadmap, and if you build something, tell us.