The rapid development of artificial intelligence (AI) is driving demand for computing power to unprecedented heights, but the underlying infrastructure faces profound structural challenges. The monopoly and high costs imposed by centralized cloud providers, along with the massive resource waste inherent in existing decentralized networks, have become bottlenecks to industry development. Gonka was born to address this core contradiction, dedicated to building a truly efficient, decentralized AI computing network.
Gonka’s core vision is to ensure that every bit of computing power is invested in meaningful AI tasks, rather than wasted on meaningless hashing or staking. In this ecosystem, hosts earn revenue by contributing GPU computing power; developers are freed from the constraints of centralized cloud services and closed-source LLMs ; and users directly create network demand by initiating AI inference requests.
Core advantage: Subverting the traditional proof-of-work mechanism and maximizing AI computing efficiency
Gonka’s innovation lies in its radical reimagining of the traditional Proof-of-Work (PoW) mechanism. We firmly believe that PoW is the only consensus mechanism that can incentivize continuous improvement and efficiency gains in infrastructure, as demonstrated by Bitcoin’s history. However, traditional PoW consumes computing power solely for network security, which is undoubtedly a massive waste.
Gonka’s “Sprint Consensus Protocol” perfectly addresses this pain point. It doesn’t simply replace hashing with AI computing; instead, it deeply integrates network security with productive tasks. During a time-limited “sprint” phase, hosts demonstrate their computing power contribution by executing a Transformer model with 20 billion parameters. Voting weight is determined entirely by computational effort, ensuring the fair principle of “one unit of computing power, one vote.”
This makes Gonka fundamentally different from other decentralized computing networks:
Compare this to Bittensor: In Bittensor, up to 60% of rewards go to stakers who don’t contribute to the hashrate, resulting in hashrate providers (such as Subnet 64, which contributes the most advanced GPUs) receiving only a tiny fraction of the network rewards. Gonka’s model, on the other hand, distributes nearly all rewards directly to hashrate contributors, eliminating this unfair resource misallocation. Compare this to Akash Network: Akash’s rewards largely go to stakers, failing to effectively subsidize AI computing infrastructure. Gonka draws on the economic incentives of Bitcoin’s PoW system, reducing AI computing costs by 90% or more through block subsidies, thereby continuously incentivizing the GPU hardware ecosystem to move towards higher efficiency and performance.
In addition, Gonka uses probabilistic verification and a reputation system, reducing verification overhead to 1-10%, while maintaining strong network security and fundamentally eliminating the waste of computing power.
Top endorsement: team and ecosystem
The launch and development of the Gonka project wasn’t entirely without foundation; it’s backed by a strong founding team and industry ecosystem. The project was incubated by US-based AI developer Product Science Inc. , founded by the Libermans siblings , Web 2 industry veterans and former core product directors at Snap Inc.
Product Science has been a pioneer in the AI field since 2021, with clients including industry giants such as Walmart, JPMorgan Chase, and Airbnb . In 2022, the company successfully raised $18 million from a stellar investor list, including a16z founders Marc Andressen and Chris Dixon, Arielle Zuckerberg, and Notion founder Akshay Kothary . This recognition from top investors and industry leaders has laid a solid foundation for Gonka’s long-term development.
Early contributors to the project include renowned leaders in the Web 2-Web 3 space, including 6 blocks , Hard Yaka , Gcore , and Bitfury . Gonka also boasts a strong advisory team, including Greg Kidd of Hard Yaka, Patrick Dai of Qtum, and Val Vavilov of Bitfury, providing valuable insights into its strategic development.
Economic Model and Fair Launch: GNK Token
Gonka’s native token, GNK, represents the future of AI computing power, a core asset. Its total supply is fixed at 1 billion and follows a Bitcoin-like issuance schedule, halving roughly every four years. This scarcity-driven economic model aims to create long-term, sustainable value for the network.
Gonka adheres to the principles of a fair launch. The only way to obtain GNK tokens is to contribute GPU computing power to the network (mining) or purchase them from the community pool. “The future direction of the network will be entirely determined by the community,” the team stated.
The Gonka testnet is now live and growing, with connected GPU computing power increasing by 50% monthly. The project has also passed the Certik audit. Gonka is redefining the future of decentralized computing with its unique model, providing a solid infrastructure for the next leap forward in the AI industry.
Gonka is a next-generation decentralized AI infrastructure designed to fundamentally address the inefficiency of blockchain computing power and make advanced AI accessible to everyone. Leveraging a unique Transformer-based PoW mechanism, Gonka converts 100% of GPU computing power into meaningful AI training and inference tasks, empowering developers to build applications freely while ensuring fair returns for computing power contributors. Gonka was incubated by Product Science Inc. , a pioneer in the US AI field. Early contributors include renowned companies such as 6 Blocks, Hard Yaka, Gcore, and Bitfury. Its advisory team includes Greg Kidd of Hard Yaka, Patrick Dai of Qtum, and Val Vavilov of Bitfury .
Official Website | Github | X | Discord | Whitepaper | Economic Model | User Manual