Testnet Live on Arbitrum Sepolia

Closed-Loop Scientific Discovery at Scale

Test 10,000 parameter combinations in parallel. Verify results on-chain. Close the loop with physical experiments — all without trusting a single centralized server.

10K+ Batch Inputs
3+ Node Consensus
99.99% Gas Savings

Why OARN?

Built from the ground up for decentralization, privacy, and open science

No Hardcoded Values

All infrastructure discovered dynamically via ENS and DHT. No central points of failure.

Wet Lab Oracle Live

Physical experiment results recorded on-chain. Multiple certified labs confirm each measurement. The only AI compute network that closes the loop with reality.

Built on Arbitrum

Low fees, fast finality, Ethereum security. Perfect for microtransactions.

Batch Processing

Submit 10,000+ parameter combinations in a single task. Nodes execute in parallel with 99.99% gas savings.

Community Governed

GOV token holders decide protocol upgrades, fee structures, and treasury allocation.

Earn COMP Rewards

Run a compute node and earn COMP tokens for every task you complete.

How It Works

From hypothesis to verified results — on-chain

1

Submit Batch Task

Upload your model and a parameter grid — up to 10,000 combinations. One on-chain submission, 99.99% gas savings vs individual tasks.

2

Parallel Execution

Independent nodes claim and execute your task. Each runs the full parameter sweep and submits a result hash. No coordination, no central server.

3

On-Chain Consensus

Nodes that agree earn rewards. Nodes that deviate earn nothing. The majority hash is the verified result — auditable on Arbitrum forever.

4

Wet Lab Validation Live

Top predictions queue for physical experiments. Lab results are submitted to the WetLabOracle contract. Each validated round trains a better model for the next.

Why the Loop Matters

Consensus proves execution. Correctness is a different problem.

The Problem

Most AI systems maximize iteration speed.
Nobody closes the loop with physical reality.
Consensus proves execution — not correctness.
A model can deterministically produce a wrong answer and 3 nodes will agree on it perfectly.

Our Solution

  • Compute — 10,000 combos in minutes via OARN nodes (live)
  • Verify execution — multi-node consensus on Arbitrum (live)
  • Verify correctness — WetLabOracle, physical results on-chain (live)
  • Self-improving — each round: better model, fewer tests needed

Web Dashboard

Access the network through an intuitive interface - no command line required

Node Operator

  • Browse & claim available tasks
  • Submit computation results
  • Track earnings with charts
  • View wallet balances

Researcher

  • Submit single inference tasks
  • Batch submit 10,000+ parameters
  • Track consensus status
  • Download verified results

Crowdfunder

  • Browse fundable research
  • Contribute ETH to tasks
  • Track funded research
  • See impact metrics

Investor

  • Network-wide analytics
  • Token metrics & charts
  • Governance participation
  • Vote on proposals
Launch Dashboard View Source

Built with Next.js 14, RainbowKit, and Tailwind CSS

Dual Token Model

Designed for sustainability and fair value distribution

GOV Governance Token
  • Supply: 100M (Fixed)
  • Purpose: Voting, staking for providers
  • Distribution:
    • 40% Early contributors
    • 30% Public sale
    • 20% DAO treasury
    • 10% Core team (4yr vest)
COMP Compute Token
  • Supply: Inflationary (decreasing)
  • Purpose: Task rewards, payments
  • Emission:
    • Year 1: 100M
    • Year 2: 80M (-20%)
    • Year 3: 64M (-20%)
    • Optional 2% burn on transfer

Earn GOV Before Mainnet

Run a testnet node now. Every task you complete earns points that convert to a guaranteed GOV allocation at TGE (Q3 2026).

Bronze
100+ pts
1,000 GOV

Complete ~7 consensus tasks

Silver
500+ pts
5,000 GOV

~33 tasks + weekly uptime bonus

Gold
2,500+ pts
25,000 GOV

~160 tasks + referrals + bonuses

Platinum
10,000+ pts
100,000 GOV

Top operators + contributions

Genesis
Top 10 nodes
250,000 GOV

Top 10 operators by total points

Ways to Earn Points

+15 Task completed with consensus
+50 100% uptime for a full week
+500 Alpha tester (first 20 operators)
+100 Valid bug report accepted
+200 Refer a node operator (2+ weeks active)
+150 GitHub PR merged

10M GOV reserved for testnet operators. Leaderboard updates every Monday. Final snapshot before Q3 2026 TGE.

Live on Testnet

Deployed on Arbitrum Sepolia - Try it now!

OARNRegistry 0xa122518Cb6E66A804fc37EB26c8a7aF309dCF04C View on Arbiscan
TaskRegistryV2 0xD15530ce13188EE88E43Ab07EDD9E8729fCc55D0 View on Arbiscan
WetLabOracle 0xF8991A56cB5B9073a3eEC87E95Dfb055fdDF0094 View on Arbiscan
GOV Token 0xB97eDD49C225d2c43e7203aB9248cAbED2B268d3 View on Arbiscan
COMP Token 0x24249A523A251E38CB0001daBd54DD44Ea8f1838 View on Arbiscan

Get Started

Choose your path

Run a Node

Contribute compute power and earn COMP rewards

git clone https://github.com/oarn-network/oarn-node
cd oarn-node
cargo build --release
./target/release/oarn-node start
View on GitHub

Submit Batch Tasks

Run thousands of parameter combinations in parallel

npm install @oarnnetwork/sdk

import { OARNClient, generateParameterGrid }
  from '@oarnnetwork/sdk';

const client = new OARNClient({
  privateKey: process.env.PRIVATE_KEY
});

// Generate 10,000 parameter combinations
const inputs = generateParameterGrid({
  temperature: { min: 20, max: 40, steps: 100 },
  concentration: { min: 0.1, max: 1.0, steps: 100 }
});

// Submit as single task (99.99% gas savings)
const { taskId } = await client.submitBatchTask(
  modelBuffer, inputs,
  parseEther('0.1'), 5, deadline
);

Join Governance

Help shape the future of decentralized AI

  • Acquire GOV tokens
  • Delegate or self-delegate voting power
  • Vote on proposals
  • Create proposals (with sufficient stake)
Join Community

Join the Community

Connect with researchers, developers, and node operators

Support the Project

Help us build open AI infrastructure for everyone