Crynux operates a permissionless, trustless network that distributes AI inference, fine‑tuning, and training workloads across a global pool of edge GPUs and Apple Silicon devices. Developers access these capabilities through OpenAI‑compatible APIs, while node operators earn CNX tokens for completed tasks and model owners tokenize assets for automated revenue sharing.
Funding
Funding not disclosed
Founders
Product
Problem
AI model inference, fine‑tuning, and training are typically confined to centralized cloud providers or require local high‑end GPU hardware, limiting accessibility, increasing latency, and creating trust and cost concerns for developers and edge device owners.
Solution
Crynux operates a truly permissionless, trustless network that orchestrates AI workloads across a global pool of edge devices—ranging from desktop GPUs to Apple Silicon mobiles—using a zero‑knowledge proof‑based consensus protocol (vssML) with Verifiable Random Functions. The platform exposes all AI services (LLM, Stable Diffusion, MusicGen, video diffusion, etc.) through OpenAI‑compatible APIs, so developers can run inference, fine‑tune, or train models without provisioning any hardware. Consensus overhead is limited to ~10%, delivering performance comparable to traditional cloud services while guaranteeing computation correctness even under adversarial conditions. Model and data assets are tokenized on‑chain, enabling automated revenue sharing with model owners and incentivizing node operators with CNX tokens. A lightweight SDK and Crynux Bridge simplify integration, allowing applications to consume AI capabilities with a single API call.
Target Audience
AI developers and enterprises that need scalable, low‑latency model inference or training without owning GPU infrastructure, and edge device owners or cloud providers seeking to monetize idle compute resources.
Features
- vssML consensus built on Zero‑Knowledge Proofs and VRF, providing provable correctness and Sybil‑attack resistance.
- Permissionless node onboarding (no whitelist or sign‑up); any GPU‑enabled device can join and contribute compute.
- Edge compute pool currently exceeding 47,000 TFLOPS across ~1,900 nodes, supporting models from 405 B parameters to n‑bit quantized variants.
- OpenAI‑compatible endpoints via the Crynux Bridge for LLM, image, audio, and video generation, with support for tool‑calling, streaming, and structured output.
- On‑chain model tokenization and data‑asset protection using federated learning and Multi‑Party Computation, enabling fair revenue distribution to token holders.
- Cross‑chain asset interoperability, allowing AI assets to be used in existing DeFi protocols.
- Crynux SDKs (Python, JavaScript, TypeScript) and Dockerized node deployments for rapid integration and scaling.
- Incentive layer that rewards node operators with CNX tokens for completed tasks, aligning network health with economic incentives.