Yeliz Akmaca 
Gonka is a decentralized platform designed to run artificial intelligence (AI) computations across a global network of GPUs. It provides an accessible and low-cost infrastructure for AI model training and inference. Developed as an alternative to traditional cloud providers like AWS or Google Cloud, Gonka focuses on minimizing resource waste.
Launched in September 2025, the Mainnet now supports large models such as Qwen3-235B. Gonka operates as a Layer 1 blockchain network, and contributors are rewarded with GNK tokens.
The Vision of the Liberman Siblings and Investor Support
The project was founded by the Liberman siblings (David, Daniil, Anna, and Maria Liberman). The siblings previously sold a company to Snapchat for $60 million and founded Product Science, providing AI code optimization services for enterprises. They bring extensive entrepreneurial experience to Gonka.
Investors and Strategic Partners
Gonka has raised approximately $69 million in funding by the end of 2025. Notable investors include:
-
Bitfury: $50 million investment in December 2025
-
Coatue Management: Supports the project with OpenAI expertise
-
Slow Ventures, K5, Hard Yaka: Additional investors strengthening the ecosystem
The funding supports network development and GPU provisioning.
Project Concept
Gonka’s core idea is to decentralize AI computation, making it both accessible and efficient. While most computing power in existing networks (e.g., Bitcoin mining) is wasted on meaningless consensus tasks, Gonka directs nearly all resources toward productive AI tasks, such as training and inference.
The project implements “Proof of Work 2.0”, maintaining blockchain security while providing developers with censorship-resistant, low-cost access. The goal is to make AI “unstoppable like Bitcoin” and to break the monopoly of tech giants such as OpenAI and Google.
How Gonka Works: Proof of Learning
Gonka operates via the Sprint mechanism. Every 10 minutes, Hosts solve Transformer-based tasks. Those producing results close to the target vector earn voting power and GNK rewards. During periods of low demand, dummy tasks keep computational resources engaged.
Training is conducted with DiLoCo distributed learning, where large model segments are shared across Hosts. Each Host trains its segment, enabling scalable and efficient network-wide AI computation. During each Sprint, participants attempt to find appropriate vectors using nonces, similar to Proof-of-Work, but the goal is AI model contribution, not block creation. Successfully generated vectors translate into GNK rewards and governance rights.
Security is ensured using random number generation, Threshold Cryptography, and Commit-Reveal mechanisms. Malicious activity is detected with state hashes and probabilistic validation, making the network secure and fair.
Governance
Gonka governance relies on Proof of Compute (PoC) weight. Each Host has a default 20% voting power, while the remaining 80% is activated through GNK collateral locking.
Decisions follow defined rules:
-
Quorum: 33.4% participation
-
Majority: >50% approval
-
Veto: 33.4% disapproval
All proposals are on-chain, usually requiring collateral. The first 180 epochs are exempt from collateral requirements.
Roadmap 2026
Gonka plans to expand its network and ecosystem integration throughout 2026:
-
Q1 2026: Testnet launch and first Proof-of-Learning Sprints
-
Q2 2026: Mainnet launch and GNK token distribution completion
-
Q3 2026: Sharding implementation and large model support
-
Q4 2026: DApp integration and ecosystem partnerships expansion
GNK Token Utility
GNK tokens are used for:
-
Rewards and incentives within the network
-
Governance and voting rights
-
Service access and AI model usage fees
Token Details
-
Token Name: Gonka
-
Symbol: GNK
-
Total Supply: Fixed at 1 billion GNK
-
Standard: ERC-20 or compatible
Token Distribution
-
Core Incentive (Hosts): 680 million GNK (68%) – Epoch rewards for computation contributors
-
Community Pool: 120 million GNK (12%) – For liquidity, ecosystem initiatives, and community programs
-
Founders Allocation: 200 million GNK (20%) – Reserved for early founding team members
This distribution ensures sustainable growth while motivating both the community and founders.
Ecosystem
The Gonka ecosystem includes participants, validators, developers, and investors, providing:
-
AI model training and validation
-
Reward and governance systems
-
DApp and third-party integrations
Designed as a transparent, secure, and sustainable distributed AI network.
Features
-
Distributed AI Training: Large models trained across shared resources
-
Proof-of-Learning: Measures and rewards computation contributions
-
On-Chain Governance: Transparent blockchain-based management
-
Sharding and DiLoCo: Model segments managed efficiently, enabling scalable training
-
Security and Validation: Malicious contributions blocked via state hashes, probabilistic validation, and honeypots
Gonka aims to optimize AI computational resources, but technical challenges and optimizations continue. The network’s potential will be clearer as it scales and supports larger models.
Official Links
You can freely share your thoughts and comments about the topic in the comment section. Additionally, please don’t forget to follow us on our Telegram, YouTube and Twitter channels for the latest news.