Balance

Balance is a combined AI and Web3 protocol and framework that connects AI agents, called AI Epals, to onchain applications, with social and gaming as the main scenarios. The architecture has four layers: infrastructure, comprising a chain abstraction layer, a node system, an identity system and a stablecoin; protocol, handling fan token issuance, trading and revenue sharing; platform, hosting first-party and third-party games; and application, providing AI battle reports, game NFTs and shareable achievement SBTs.

The core mechanism is chain abstraction plus distributed compute. The abstraction layer picks an execution chain by speed, cost and security so users need not understand cross-chain detail; the node system draws on idle GPUs of personal computers worldwide to train and deploy models, letting AI Epals run across several chains and routing each task type to the most suitable network, while nodes also verify, govern and share protocol revenue. AI battle reports turn match data into updatable NFTs, and SBTs record achievements that can be shared. In May 2026 the agent coordination model moved from top-down task allocation to agents competing for tasks on market signals with collaboration verified and settled onchain; since June 2026 the team has published research on treating verifiable onchain labour, rather than holdings, as the basis for reward settlement.