NewsBTC
Sep 17, 2026
6 0
Bullish

B.AI reaches 1.51 trillion daily tokens, launches global agent settlement layer

3 min

B.AI positions itself between foundation models and AI agents by integrating compute routing, decentralized settlement, and full-stack agent infrastructure. Five months post-launch, it reports daily throughput above 1.51 trillion tokens, a 71,500x increase since inception. The platform states it reached this scale in roughly 100 days, a timeline it contrasts with OpenRouter’s growth curve, attributing acceleration to integrated settlement and core infrastructure.

Infrastructure Scale and Model Distribution

B.AI aggregates international models such as GPT and Claude alongside Chinese LLMs and reports close partnerships with Tencent, Alibaba, ByteDance, DeepSeek, Zhipu AI, Kimi, and MiniMax. High-usage Chinese models on the platform include DeepSeek-V4-Flash, Qwen3.8-Flash, and GLM-5.3 Flash. B.AI uses model discounts, time-limited free access, tiered API pricing with up to 90% discounts, and developer incentives to convert model cost advantages into throughput.

YZi Labs acts as a distribution engine by onboarding early-stage projects and developer communities. B.AI targets high-throughput, low-cost inference demand, positioning as a global compute distribution hub for LLM providers seeking large, concurrent workloads.

Settlement Layer and Dual-Track Payments

B.AI embeds a settlement layer intended to meter and settle cross-region calls and task allocations. It operates a dual-track payment network: on-chain settlement channels for decentralized, verifiable, low-friction automated transactions and micropayments; and legacy Web2 gateways including WeChat Pay, Alipay, UnionPay, and Visa for enterprise compatibility. The design offers a single integration point for global capital flow.

Integration with TRON provides access to the TRC-20 USDT stablecoin network as a payment rail. As of September 10, TRC-20 USDT supply stood at approximately 94.2 billion, enabling cross-border compute scheduling and high-frequency agent micro-transactions on-chain.

Agent Runtime and Protocol Stack

B.AI builds a full-stack runtime to support automated, agent-to-agent commerce:

  • Scheduling layer: tiered API system for fine-grained compute allocation aligned with specific business demands.
  • Execution layer: native agent collaboration with embedded toolchains such as Skills and Agent Wallet, plus integrated assistants BAIclaw and BAI code to enable autonomous data acquisition and task execution.
  • Protocol layer: integrated x402 payment protocol and 8004 identity authentication protocol to standardize trust, identity verification, and high-frequency settlement between agents.

This stack links perception, reasoning, action, and settlement, targeting scalable micro-transactions and compute settlement for the agent economy.

Key Data Points

  • Daily token throughput: 1.51 trillion, up 71,500x since launch.
  • Chinese LLM partnerships: Tencent, Alibaba, ByteDance, DeepSeek, Zhipu AI, Kimi, MiniMax.
  • High-usage models: DeepSeek-V4-Flash, Qwen3.8-Flash, GLM-5.3 Flash.
  • Payments: dual-track Web3 and Web2; TRON USDT integration with 94.2 billion TRC-20 USDT supply as of September 10.

B.AI presents its infrastructure as a closed loop from compute scheduling through agent execution to automated settlement, targeting A2A collaboration and large-scale, high-frequency transactions in the agent economy.