OpenGradient

OPG is used primarily for paying for LLM inference through OpenGradient’s x402 payment flow. In that setup, payments are made on Base, while inference execution, proof settlement and verification happen on the OpenGradient network.

OpenGradient full nodes manage payment settlement and fee distribution, and x402 requests require cryptographically verified OPG payments before execution.

Beyond LLM payments, the wider network uses on-chain settlement for AI-related operations including:

• verified LLM inference
• on-chain ML execution through PIPE
• model hosting through the Model Hub
• memory operations through MemSync
• digital twin interactions through Twin.fun

The project portal publishes a token allocation page for OPG, with allocations for the ecosystem, foundation, core contributors, investors and advisors, staking rewards, liquidity and airdrop.

OpenGradient’s architecture has:

• Full nodes that run consensus, verify TEE attestations and ZKML proofs and settle payments
• Inference nodes that execute models, either through TEE proxy nodes or local GPU inference nodes
• Walrus storage for models and large proof data

The diagram on page 6 shows the split between execution and verification, the table on page 7 shows the fast path where users get results quickly and an asynchronous verification path where attestations or proofs are checked later, and pages 10 and 11 explain how TEE and ZKML verification are used.