OpenGradient

• Name: OpenGradient
• URL: https://opengradient.ai
• Category: AI-inference settlement layer / verifiable-compute network / TEE-and-ZKML-backed execution infrastructure / crypto × AI middleware
• Summary: OpenGradient is best understood not as a generic AI chain, but as a verification-and-settlement stack for AI execution. Its core mechanism is the Hybrid AI Compute Architecture (HACA), which splits fast offchain inference from onchain proof settlement: specialized inference nodes run models or proxy third-party LLMs inside TEEs, full nodes verify proofs and maintain the ledger, data nodes fetch external inputs inside enclaves, and large artifacts live offchain on Walrus. That makes OpenGradient a useful comparison point for crypto × AI systems where the real control surfaces are node registration, attestation policy, verification-method choice, payment authorization, and proof settlement rather than the broader marketing label of decentralized AI.
• What it does:
  • Positions itself as a purpose-built network for verifiable AI inference rather than as a general-purpose smart-contract chain
  • Uses HACA to separate inference execution from verification, so requests go directly to specialized compute nodes while proofs settle asynchronously onchain
  • Splits the network into full nodes, inference nodes, data nodes, and decentralized storage, with each layer handling a distinct trust or workload boundary
  • Supports a verification spectrum spanning TEE attestations, ZKML proofs, and unsigned or signature-only Vanilla execution paths depending on the workload
  • Runs an onchain registry for TEE and inference-node admission, attestation checks, revocation, and result-signing trust
  • Uses an x402-style payment flow on Base for LLM inference while keeping node registration, proof settlement, and verification on the OpenGradient network
  • Exposes developer-facing products including verifiable LLM inference, a model hub, SDKs, and agent-oriented tooling around the settlement layer
• Key claims:
  • The clearest analytical split is between execution and verification. OpenGradient’s docs repeatedly emphasize that inference should not sit in the blockchain critical path; instead, the chain exists to authorize nodes, settle payments/proofs, and preserve an auditable record of verified AI operations.
  • HACA matters because it makes the workload split explicit: GPU-heavy inference nodes execute, full nodes verify without re-execution, data nodes fetch external inputs under enclave guarantees, and Walrus stores large models/proofs. That is a cleaner control-plane decomposition than the flatter AI blockchain label suggests.
  • OpenGradient’s verification spectrum is one of its strongest reusable ideas. It does not insist every workload use the same trust model; instead it lets developers choose TEE, ZKML, or lighter-weight verification depending on whether they prioritize latency, cryptographic certainty, privacy, or cost.
  • The TEE registry and node-registration path are important because they show where practical trust sits: not in generic decentralization rhetoric, but in attestation policy, approved code images, revocation, and which node signatures full nodes will accept.
  • The x402-on-Base plus proof-settlement-on-OpenGradient split is analytically useful because it separates machine-payment authorization from AI-verification state. That makes OpenGradient a strong bridge between the corpus’ x402 work and the crypto × AI execution cluster.
  • OpenGradient cleared the bar for the active corpus because it adds a distinct comparison point for verifiable AI middleware: a chain whose primary job is not broad app execution but the settlement, verification, and registry layers around AI inference.
• Whitepaper: OpenGradient published a March 2026 whitepaper, ../whitepapers/opengradient-whitepaper.pdf, alongside official architecture and inference docs; see ../whitepapers/opengradient-primary-sources-2026-05-12.md.
• Sources:
  • https://opengradient.ai
  • https://docs.opengradient.ai/about/
  • https://docs.opengradient.ai/learn/architecture/
  • https://docs.opengradient.ai/learn/network/consensus.html
  • https://docs.opengradient.ai/learn/onchain_inference/llm_execution.html
  • https://opengradient.foundation/whitepaper
  • https://opengradient.foundation/opengradient-whitepaper.pdf
  • https://github.com/opengradient

Internal linkages

• Best upward reads: gensyn and ritual.

Control surface

• The leverage sits in node registration, attestation and revocation policy, proof-mode choice, payment authorization, and the boundary between fast offchain inference and what the chain is willing to accept as settled execution.

• Treat it as AI-verification and settlement middleware, not as a general-purpose AI chain.

• Last reviewed: 2026-06-14 UTC