Zero-Knowledge AI Verification
Deploy verifiable AI models off-chain and settle cryptographic proofs on-chain in milliseconds. Integrate trustless execution into your protocol using our lightweight developer toolkit.
Three Lines of Code
Our zero-knowledge protocol proves that a specific model generated a specific output. Generate cryptographic proofs locally and verify them on-chain with minimal latency.
$ npm install @inference-labs/sdk $ node verify.js [info] Initializing proof generation... [info] Model: llama-3-8b-instruct [success] Proof generated: 0x9f3a...8a21 [success] Settle on-chain: Verified
Multi-Language SDKs
Native libraries designed for high-performance cryptographic operations. Spin up local testnets and generate zero-knowledge proofs in your preferred runtime environment.
Rust SDK
Go SDK
TypeScript SDK
High-performance proof generation for native validators and node operators.
Robust consensus integration for custom blockchain clients and decentralized networks.
Lightweight client-side verification for web applications and decentralized frontends.
cargo add inference-sdk
go get github.com/inference
npm install @inference-labs
Protocol Specs
Have complex cryptographic requirements? Read our architectural specs or connect with our engineering team directly on our integration channels.
How does Proof of Inference protect model weights?
What is the latency of on-chain verification?
Is it compatible with EigenLayer and Bittensor?
The model runs inside a secure enclave or zero-knowledge prover. Only the mathematical proof of correct execution is published on-chain, keeping weights completely private.
Proofs are generated off-chain in milliseconds. On-chain verification is highly optimized, settling in a single transaction block on supported networks like EigenLayer.
Yes. We operate active subnets on Bittensor Subnet 2 and offer testnet integrations on EigenLayer, allowing validators to easily verify off-chain compute.