B.AI, an AI infrastructure platform, reported crossing 1.33 trillion daily tokens in throughput following the launch of free access to top-tier models. The milestone was reached within days of the rollout, marking a significant adoption surge across the platform.
Over a 15-day period, cumulative token volume reached 8.19 trillion, drawing in more than 220,000 new API users. As of September 3, B.AI's total user base had surpassed 2.3 million users.
The platform offers unlimited free access to six frontier models: DeepSeek-V4-Flash, DeepSeek-V4-Flash-Vision-Exp, Tencent Hy3, Xiaomi MiMo-V2.5, GLM-5.3-Flash (Ox Alpha), and Qwen3.8-Flash. On September 3, B.AI introduced a new pricing structure for DeepSeek-V4-Flash models, offering a 50% discount during peak hours and 25% of standard peak pricing during off-peak hours, while maintaining free access to the other four models.
Full-Stack Infrastructure for AI Agents
B.AI positions itself as a settlement layer for AI agent workflows, integrating diverse models under a unified infrastructure. The platform operates a dual-tier API structure offering both official-route reliability and discounted custom channels, allowing developers to select between guaranteed connections and cost-optimized options.
The platform supports both traditional Web2 payment methods and Web3 on-chain payment systems, enabling developers to choose their preferred settlement approach.
Core Infrastructure Components
B.AI's infrastructure includes five core components designed to support agent operations:
- x402 Payment Protocol: Executes high-frequency, on-chain micro-settlements during cross-agent API calls and compute orchestration.
- 8004 Identity Protocol: Issues verifiable on-chain credentials for agents and logs execution history.
- Skills Matrix: Standardized building blocks that interface with MCP servers for tool-calling capabilities.
- BAIclaw and BAIcode: Platform assistants for terminal operations, multi-agent workflows, and end-to-end development tasks.
The platform also provides compatibility with Codex, allowing developers to use B.AI with GPT models and DeepSeek models within their development workflows.
