OpenGradient is a decentralized AI inference network designed to enable verifiable AI computation. It separates execution from verification, using technologies like TEE and ZKML to ensure the trustworthiness of inference results while maintaining Web2-like latency. Developers can call LLM or ML models via SDK, or integrate AI functionality directly into smart contracts. Its mechanisms include the HACA architecture, CometBFT consensus, and the OPG token economy.