B.AI Launches Global Settlement Layer, Hits 1.51 Trillion Tokens

Singapore-based B.AI launched a global settlement layer and full-stack agent infrastructure and recorded daily throughput of 1.51 trillion tokens five months after launch, the company reported. The platform said traffic has grown more than 71,500-fold since its first day and that it reached trillion-token scale in just over 100 days.

B.AI was co-incubated by TRON and YZi Labs. The platform aggregates international models such as GPT and Claude alongside Chinese models from DeepSeek, Kimi, Qwen and GLM. The company has partnerships with Tencent, Alibaba, ByteDance, DeepSeek, Zhipu AI, Kimi and MiniMax. Models with the highest usage on the platform include DeepSeek-V4-Flash, Qwen3.8-Flash and GLM-5.3 Flash.

The platform architecture has three functional layers. The scheduling layer offers fine-grained compute allocation and structured invocation plans to match workloads to cost and performance needs. The execution layer provides a native agent runtime with toolchains called Skills, an Agent Wallet, and built-in assistants named BAIclaw and BAIcode to support data acquisition and task execution. The protocol layer implements payment and identity standards referenced as the x402 payment protocol and the 8004 identity protocol for metering and automated settlement.

B.AI supports a dual-track payments model that remains compatible with Web2 gateways including WeChat Pay, Alipay, UnionPay and Visa while also offering on-chain settlement options. The platform uses the TRON network and its TRC-20 USDT ecosystem as a native payment rail. The company cited TRON network data showing TRC-20 USDT supply surpassed 94.2 billion as of Sept. 10.

YZi Labs functions as a distribution engine for the B.AI ecosystem, onboarding early-stage projects and developer communities and driving use cases that increase platform traffic. To accelerate adoption, the platform offers tiered API pricing, time-limited free access and discounts the company describes as up to 90%.

The company said the combination of model aggregation, compute routing, embedded settlement and an agent runtime is intended to support frequent, small payments between software agents and to meet demand for high-throughput, lower-cost inference. B.AI presented its settlement layer and protocol stack as components for automated agent workflows that require metering and cross-region settlement.