Single-day token throughput exceeds 1.33 trillion: B.AI builds the "AI Grid" with full-stack infrastructure, solidifying the "global settlement layer" of the intelligent economy

Recently, the new generation AI infrastructure platform B.AI launched the "Universal Access to Large Model Computing Power" initiative, sparking an unprecedented surge in large model calls among developers across the internet. Within just a few days, B.AI's platform achieved a historic daily token throughput of over 1.33 trillion, with daily API calls exceeding 10.86 million. In the past 15 days, B.AI's core throughput experienced an explosive growth of 289 times.

The achievement of this milestone data marks the full ignition of B.AI's "Universal Access to Computing Power" initiative, but this is merely the first step in B.AI's strategic blueprint. B.AI's ultimate strategic vision is to completely transcend the simple distribution channel of computing power and to comprehensively build the "Intelligent Global Settlement Layer" as a core hub, thereby truly supporting the cross-node collaboration, scheduling, and value distribution of massive agents in complex businesses.

As the AI industry fully enters the era of high-frequency agent collaboration, facing an increasingly large demand for concurrency, major underlying large models are like isolated "power plants" scattered across various locations, while B.AI becomes the "AI grid" that connects the entire era. B.AI, with its forward-looking vision, accurately positions itself "above all models and below all agents." By deeply integrating diverse top models and full-stack components, it has laid a solid, irreplaceable infrastructure foundation for the large-scale implementation and productivity explosion of agents.

Daily Throughput Exceeds 1.33 Trillion! B.AI's "Universal Access to Computing Power" Event Ignites a Frenzy of Calls Across the Internet

Recently, the free usage event for large models launched by B.AI quickly fermented within the developer community, becoming an unprecedented focal topic in the industry. This initiative, which can be described as a "computing power carnival," not only brought considerable activity to the platform but also set a series of impactful high-profile data records.

In just a few days, the call volume on the B.AI platform surged, driving the entire platform's daily token throughput to strongly exceed 1.33 trillion, with a cumulative token throughput of 8.19 trillion over 15 days, and over 220,000 new API users. As of September 3, B.AI's total platform users surpassed 2.3 million. This traffic surge triggered by "Universal Access to Computing Power" is refreshing the usage records of AI infrastructure with the most genuine market response.

Such a scale of calls is directly attributed to B.AI's sincere and precise targeting of developers' "cost anxiety" pain points with its free usage matrix. In this event, B.AI fully opened up 6 leading top models, providing the entire network with a "0 threshold, unlimited " usage experience.

This production-grade matrix accurately covers the full scene demands of advanced applications: GLM-5.3-Flash (Ox Alpha) and DeepSeek-V4-Flash-Vision-Exp strongly support multi-modal vision and ultra-long text parsing; DeepSeek-V4-Flash collaborates with Qwen3.8-Flash, catering to ultra-high concurrency responses and complex code logic generation; paired with Tencent Hy3, which is deeply rooted in the Chinese base, and Xiaomi MiMo-V2.5, which adapts to all scenarios. From rapid text and code reasoning to advanced visual analysis, B.AI has completely cleared the cost barriers for developers to access complex AI applications with this top-tier lineup.

It is this "zero threshold" top model matrix that has directly catalyzed the surge in developers' usage of B.AI. In the past, high token bills directly limited the implementation of complex agent architectures. However, when B.AI completely leveled the computing power costs, developers could finally invest those agent architectures capable of "self-planning, tool invocation, reflection, and error correction" for long-chain tasks on a large scale into real production and testing environments.

It is worth noting that starting from September 3, B.AI adjusted the pricing standards for the two models, DeepSeek-V4-Flash and DeepSeek-V4-Flash-Vision-Exp, introducing a tiered discount (50% off during peak hours, as low as 25% of the official peak price during off-peak hours), while retaining the free mechanism for the four large models: GLM-5.3-Flash (Ox Alpha), Qwen3.8-Flash, Hy3, and MiMo-V2.5. In the face of this commercialization transition, developers' enthusiasm for calls remains undiminished, and the platform's token throughput continues to maintain a strong upward momentum.

This fully demonstrates that this wildly popular event is not only a concentrated release of computing power dividends but also an excellent "barometer" for the evolution of industry infrastructure. It clearly indicates with a daily throughput of 1.33 trillion tokens that AI applications have officially bid farewell to the simple "Q&A era," and B.AI is comprehensively opening up deep empowerment for the large-scale implementation of advanced AI agents with a solid technical foundation and refined service mechanisms.

B.AI Builds the "AI Grid," Solidifying the "Global Settlement Layer" of the Agent Economy

The "Universal Access to Computing Power" created by B.AI is just the starting point, not the endpoint. For B.AI, this "computing power carnival" is also an extreme stress test of the platform's underlying capabilities. When massive developers flood in at the same time, and long-chain agent tasks circulate on the platform with unprecedented density, the real question arises: how to ensure that this explosion is not just a one-time event but solidifies into a sustainable, evolvable, and settleable systemic capability? Standing at this juncture, B.AI chooses to focus its strategic vision deeper, committed to building a full-stack infrastructure that supports the operation of the agent era, solidifying the positioning of the "Intelligent Global Settlement Layer."

In traditional understanding, the significance of infrastructure lies in "stable supply." However, in the agent era, the connotation of infrastructure is undergoing a qualitative change. A typical agent task often requires switching repeatedly between multiple models: using visual models to understand input, using reasoning models to plan steps, using code models to generate executable logic, and then using lightweight models to complete formatted output. This "multi-model relay" work mode determines that no single model vendor can independently support a complete agent workflow. Developers are forced to face the fragmented reality of multiple API protocols, multiple billing systems, and multiple rate-limiting rules.

The settlement layer built by B.AI precisely bridges this gap. B.AI, with its forward-looking vision, accurately positions itself "above all models and below all agents," allowing models from different vendors, with different capabilities and cost structures, to be abstracted into a schedulable resource.

Relying on the "official stability guarantee, self-selected bottom price" tiered API system, developers can obtain the certainty of direct connection from the official side while flexibly choosing high-quality supplies with different discount gradients from a rich array of models, completing tasks of equal quality at more competitive costs. Combined with the "intelligent routing" function on the Chat side, B.AI has created a full-stack "AI Grid," allowing every task request from agents to land precisely on the most suitable model with reasonable costs and stable performance.

Moreover, the settlement capability of this grid also accommodates the dual logic of Web2 and Web3. On the Web2 side, developers can use familiar traditional payment methods to recharge and bill with extremely low thresholds, seamlessly integrating into existing business processes; on the Web3 side, B.AI relies on on-chain payment channels to provide global developers with decentralized, verifiable, and low-friction payment options.

The parallel integration of these two payment systems allows developers and agent applications with different backgrounds and infrastructure preferences to find the most suitable settlement path on this grid, truly achieving "one-time access, global circulation."

Directly Addressing Core Productivity, B.AI's Full-Stack Infrastructure Reshapes the Agent Collaboration Network

After establishing a solid computing power routing and global settlement foundation, B.AI further completed the leap from infrastructure laying to productivity reshaping, fully empowering agents to achieve efficient collaboration in diverse business scenarios, filling in the most critical execution layer puzzle for the comprehensive explosion of the agent economy.

The core of this puzzle is the support for Codex integration. Codex, as the core tool for code execution and agent action layers, has absolute technical advantages in understanding project architecture, proactively planning, and executing complex development tasks. Currently, B.AI API is compatible with Responses API. Developers can use the** B.AI API Key to call the GPT series models supported by B.AI in Codex, as well as the highly favored** DeepSeek series models.

Through this seamless integration, technical teams can embed these two top-tier code computing powers directly into their daily coding workflows, meeting multi-dimensional needs such as code generation, logic understanding, bug debugging, system restructuring, and development collaboration, providing developers with a complete closed-loop experience from model selection to code implementation.

Completing the seamless connection between computing power and application layers is just the first step in unleashing the productivity of large models. In addition, to enable the agent economy to take root, the diverse agent collaboration barriers beneath the task chain must also be addressed. To ensure that massive agents can operate stably in real production environments, B.AI has built a full-stack infrastructure consisting of five core components, equipping agents with an extremely precise operational system:

l x402 Payment Protocol: Unlocks automatic micro-payments between agents. Built on the HTTP 402 standard, this protocol innovatively introduces a "response prepayment" mechanism, allowing for seamless execution of high-frequency on-chain automatic micro-payments and settlements in the background when API calls or computing power collaborations occur across agents.

l 8004 Identity Authentication Protocol: Creates a dedicated "digital ID" for agents. The 8004 protocol generates objective and verifiable identity credentials for each agent on-chain, accumulating their call history and credit scores, completing the key credit loop for cross-node collaboration in the intelligent economy.

l "Skills" Skill Blocks: Provides agents with an out-of-the-box toolchain, deeply integrated with the MCP Server. The standardized Skills matrix endows agents with highly modular tool invocation capabilities, thoroughly bridging the last mile from "thinking and decision-making" to "action execution."

l BAIclaw and BAIcode: The platform's native super assistants, as super application forms aimed at end users and developers, BAIclaw and BAIcode are the ultimate embodiment of B.AI's underlying architecture and scheduling capabilities. Among them, BAIclaw focuses on omnipotent intelligent terminals and multi-agent collaboration, deeply integrating top large model matrices, allowing users to input natural language commands to automatically drive agents to complete complex on-chain economic activities and all-day hosting. BAIcode, as an advanced AI programming tool, can leverage the top intelligence of large models to achieve a complete closed-loop process from requirement understanding, architecture planning to code generation and testing.

From the surge in token throughput triggered by the free usage carnival to the establishment of the architecture as the "Intelligent Global Settlement Layer"; from seamless integration with Codex to the full-stack infrastructure built by x402, 8004, Skills, and native assistants, B.AI demonstrates not only a hardcore technical solution but also a clear vision for the future. In this AI-driven industrial reshaping, B.AI firmly positions itself "above all models and below all agents," transforming the vast and scattered supply of computing power into a highly available, transferable, and settleable system-level network capability.

As millions of agents seamlessly flow, collaborate freely, and settle in milliseconds on this infrastructure, AI is accelerating its departure from the one-way response tool attribute, truly evolving into an economic entity that drives the leap in social productivity. What B.AI has solidified is not only the present of universal access to computing power but also the ultimate foundation for the large-scale explosion of the agent economy. The future has arrived, and this is just the beginning.