Dashboards
Markets
Indicators
ETF Flows
Derivatives
Exchanges
Centralized Exchanges
Decentralized Exchanges
Crynux (crynux.io) priceCNX
This is a preview page.
For more details on listing tiers, refer to Listings Review Criteria Section B - (3).
For more details on listing tiers, refer to Listings Review Criteria Section B - (3).
Do you own this project?
- Total supply
- 8.61B CNX
- Self-reported circulating supply
- 592.07M CNX
Max. supply
8.61B CNX
Crynux (crynux.io) community
Crynux (crynux.io) Holders
Loading...
Top Holders
Crynux Network is a decentralized AI compute network that turns GPUs on everyday edge devices into a shared resource for running inference and fine-tuning tasks on large language models and vision-language models. Its first mainnet, called the Lithium Network, went live on June 17, 2026. From that point forward, applications could access these compute resources through a permissionless network of nodes.At the heart of the system is the vssML consensus protocol, which relies on verifiable secret sampling to keep the network both secure and efficient. When a task arrives, the blockchain creates a sampling seed. The submitting application uses this seed with a Verifiable Random Function to generate a random number that decides whether the task will be validated. Only a small share of tasks—around 10 percent on average—are selected. Those chosen tasks are duplicated into three identical versions that appear unrelated on the blockchain. Nodes complete the work and submit their results before learning if validation will take place. Afterward, Zero-Knowledge Proofs check that the same parameters were used across all three versions, and similarity hashes of the outputs are compared. Inconsistent results trigger penalties, including loss of fees or slashing of staked tokens. This selective approach reduces unnecessary computation while still catching dishonest behavior.Anyone can run a node by installing the open-source software on compatible hardware. Windows systems require an NVIDIA GPU with at least 8 GB of VRAM, while Apple Silicon Macs with M1 processors or newer are also supported. Nodes can also operate through Docker or on third-party cloud GPU providers. Operators earn CNX tokens for the compute they provide. A staking score, calculated from both self-staked and delegated CNX, determines how frequently a node receives tasks. This mechanism favors operators who commit more capital and maintain strong performance.Applications interact with the network through an OpenAI-compatible Bridge API and a lightweight SDK. These interfaces make it easy to connect existing tools such as LangChain, LangGraph, and agent frameworks like Hermes Agent. The network currently handles text generation with large language models, vision-language model tasks, image generation, and model fine-tuning. Developers can run fine-tuning jobs on the network and then offer the resulting models as paid services.CNX acts as the native token for fees and rewards. Holders can delegate their tokens to node operators through the Crynux Portal without running hardware themselves. In return, they receive a share of task revenue based on the operator’s performance and staking score. Model developers can also tokenize their trained models as on-chain assets. Revenue generated by these models is distributed to token holders, and the tokens themselves can serve as collateral in decentralized finance applications. These applications may run on the Crynux blockchain or on modular EVM-compatible Layer 2 chains connected through cross-chain messaging. Bridging further extends compatibility with DeFi protocols on other blockchains.The architecture combines on-chain records with Zero-Knowledge Proofs for validating task parameters and results, along with privacy-preserving computation techniques. Encrypted data and outputs are made available through Merkle proofs recorded on-chain, while the full data remains off-chain. All core software components are open source and published on GitHub.In this way, Crynux Network delivers a permissionless compute layer for AI workloads while also enabling model ownership and on-chain economic participation grounded in verified execution.



















