On August 24, the next-generation AI infrastructure platform B.AI reached a landmark milestone as its cumulative token throughput for the fully free platform access campaign surpassed 2 trillion. This remarkable achievement stems from B.AI’s recent series of industry-leading initiatives designed to promote widespread accessibility.
Amid industry-wide cost anxiety triggered by price hikes from some leading models, B.AI is pushing forward against the tide. Since August 17, when B.AI announced the limited-time free access to DeepSeek V4 Flash—setting a daily throughput record of over 220 billion tokens—Tencent Hy3, DeepSeek-V4-Flash-Vision-Exp, and Xiaomi MiMo-V2.5 have all followed suit by opening their services fully for free. This surge in usage, fueled by the tangible benefits of high-performance computing, has not only ignited global developer enthusiasm but also vividly demonstrated the exceptional resilience of B.AI’s underlying architecture under massive concurrent loads.
This near-frenzied surge in demand reflects the most pressing reality facing the AI industry today—“cost anxiety.” As the industry fully enters the era of highly automated Agents, tokens have become the foundational currency of the new economy, with their consumption rate accelerating exponentially. Soaring inference costs have surpassed the limits of model capabilities, becoming the greatest barrier to the large-scale deployment of AI.
At this critical juncture, B.AI remains steadfast in its original strategic vision as a cross-network computing power "super-distribution hub," boldly unlocking the full potential of its underlying infrastructure. Through its innovative tiered API system—“official direct connections for stability, user-selected services for the lowest prices”—B.AI ensures enterprise-grade core business reliability while delivering unprecedented cost reductions of up to 90%. Combined with fully integrated dual-channel payment networks spanning Web2 and Web3, along with ongoing, sustained discount initiatives, B.AI is comprehensively closing the loop on the practical realization of “computing power for all.”
Shedding the "scarcity premium" of large models, B.AI is breaking through the cost barrier of AI implementation, making affordable, high-efficiency computing power a true driver of intelligent transformation across industries.
When AI implementation meets "cost anxiety," calculating computing power becomes a business survival strategy.
NVIDIA CEO Jensen Huang once made a remarkably insightful business observation: In the AI era, tokens have become a "new form of currency."
When models deeply integrate into complex business workflows, enterprises and developers no longer purchase fixed software licenses but instead pay per “intelligent output.” Each understanding, reasoning, and generation by a large model is essentially a commercial transaction consuming tokens. As the industry has fully entered the era of highly automated agents over the past year, the burn rate of this “new currency” has increased by hundreds or even thousands of times.
A Gartner study found that a typical agent workflow consumes 5 to 30 times more tokens than a traditional chatbot. This means that while users previously might have needed only dozens of model calls per day, today a simple request like “Help me complete market research” can involve hundreds or even thousands of model invocation cycles, as the agent autonomously plans, retrieves information, reflects, and generates outputs.
As demand for inference explodes exponentially, "compute costs" have become the greatest concern for all enterprises and independent developers. According to Gartner, by 2027, approximately 40% of agent projects will fail due to infrastructure cost overruns. This reveals a harsh reality: what now limits the adoption of AI is often not the inherent upper limit of model capabilities, but the persistently high inference costs.
Against this backdrop, the traditional one-size-fits-all approach to invoking large models will become completely ineffective. If developers route all Agent nodes through flagship models, the exorbitant costs will directly undermine the project’s business viability. Conversely, if cost-cutting leads to a global downgrade to inexpensive lightweight models, the Agent’s complex logic chains will frequently collapse.
The key to breaking the cost cycle lies in fully transitioning to a "multi-model collaboration" underlying architecture. This is like building a smart computing hub: directing high-complexity inference tasks precisely onto the "express lane of top-tier models," while smoothly routing basic data processing to the "affordable network of lightweight models." The market no longer needs more individual model interfaces, but rather a computing hub that combines robust underlying orchestration with scalable bargaining power.
B.AI is building a "Super Computing Hub" to reconstruct the computing power distribution network with tiered APIs and intelligent routing.
In the trend toward increasingly sophisticated demand for computing power, B.AI has officially established itself as a core pillar of the next-generation AI infrastructure. Facing the exponential rise in inference costs under Agent scenarios, B.AI has moved beyond the superficial model of traditional platforms that merely aggregate interfaces, comprehensively upgrading its strategic positioning to become a “super distribution hub” spanning global computing resources.
To address this, B.AI has introduced two API integration options—“Official” and “Selected Service Providers”—providing developers with the most cost-effective “compute resource arsenal” to fully enable the accessible, practical deployment of AGI.
First, for enterprise core production environments and highly complex inference tasks, B.AI has established an official integration channel. This channel emphasizes “direct factory API connectivity,” providing enterprises with high availability guarantees and support for the full range of models (currently 42). More importantly, leveraging significant economies of scale, B.AI’s official channel directly passes platform benefits to developers, offering baseline discounts of 10%, 15%, 30%, and even 40%. This enables enterprises to ensure absolute stability for their core operations while significantly reducing their foundational compute costs.
For non-core workflows with more flexible stability requirements, B.AI innovatively introduces a self-selected service provider channel. Under this model, developers can directly choose third-party providers such as Mix, Nebula, and OL Station, and be billed at their actual discounted rates. The platform offers seven price tiers for free selection, with the lowest price reaching as low as 10% of the original rate. Through this tiered system—“officially guaranteed stability, self-selected lowest pricing”—B.AI empowers enterprises with flexible API orchestration options, enabling them to reduce their overall AI computing costs to the optimal range.
While providing robust API support for developers at the core, B.AI also prioritizes ease of use for common enterprise business scenarios. For frontend chat interaction use cases, the platform introduces an “Auto Routing Mode” designed specifically for general users. In everyday office tasks, text analysis, and other non-API scenarios, the system intelligently analyzes each frontend prompt, dynamically matching it with the most suitable available model across the network—minimizing resource waste while significantly lowering the barrier to AI usage for non-technical users.
Remove the "scarcity premium" from large models and rebuild the underlying compute allocation logic. As developers' increasingly sophisticated scheduling strategies deeply integrate with B.AI’s tiered, high-cost-performance interfaces, the "cost barrier" that has long hindered the large-scale deployment of agents is being effectively dismantled. This is not only a significant infrastructure upgrade but also a substantial acceleration of the full commercialization of AI applications.
B.AI breaks through 2 trillion throughput in 7 days, advancing "computing power for all" through a diverse rights system.
B.AI has always upheld the core philosophy of "building a普惠 computing power hub." From continuously offering limited-time free access to cutting-edge flagship large models to constantly upgrading its API "discount channel," B.AI consistently lowers the barrier to AI adoption for businesses through tangible cost reductions, earning enthusiastic market response and deep recognition.
Recently, when DeepSeek announced an increase in model pricing, B.AI swiftly responded with a highly credible market initiative, leveraging its mature resource aggregation capabilities and ecosystem strength: On August 17, B.AI announced that DeepSeek V4 Flash would be offered for free on its platform for a limited time, while simultaneously unlocking access via both the web chat interface and API, enabling enterprises to deploy production-grade AI workflows at zero cost.
This countercyclical, inclusive initiative instantly ignited the market. Just 24 hours after the campaign launched, B.AI’s key platform metrics all hit record highs: daily token throughput surged dramatically, breaking through the 220-billion mark.
And this wave of compute democratization shows no signs of stopping. Following the overwhelming success of the first free access campaign, B.AI is building on its momentum by continuously expanding its developer benefit ecosystem. On August 21, the highly anticipated Tencent Hy3 model officially launched on B.AI and was made fully free to the public. Shortly after, on August 22, DeepSeek-V4-Flash-Vision-Exp—a model with powerful visual parsing capabilities—also joined the free initiative, further completing the puzzle of low-cost compute for enterprises in multimodal scenarios.
With the continuous expansion of free benefits, developer enthusiasm has been fully ignited. On August 24, B.AI reached a landmark milestone, as the total accumulated token throughput of the platform-wide free access campaign officially surpassed 2 trillion. This not only continues to break platform popularity records but also provides the most direct evidence of B.AI’s robust resilience and exceptional scheduling capabilities under massive scale and sustained high concurrency.
In fact, B.AI has consistently maintained an extremely sharp and frequent pace of offering benefits. Previously, B.AI already introduced groundbreaking initiatives such as “MiniMax M3 Free for a Limited Time,” “GLM 5.3 Exclusive 10% Discount,” “Qwen-3.8 MAX Free for a Limited Time,” and substantial deposit bonuses, continuously releasing platform advantages to the industry. But this is only the beginning—B.AI will continue expanding its “Universal Benefit Library,” planning more diverse model-free and major discount campaigns. The platform will always stand alongside developers, relentlessly driving down the cost barrier of AI adoption through a steady stream of computing power benefits.
This enhanced accessibility is not only reflected in free-tier benefits for individual models but has also been fully extended across core usage pathways. Recently, B.AI API’s “Official Discount Channel” underwent a major upgrade, significantly expanding its discount matrix for leading large models. Building on reliable, direct-from-vendor connectivity, B.AI now offers businesses and developers more cost-effective compute options.
At the same time, to support a global and diverse developer ecosystem, B.AI has fully integrated both Web2 and Web3 payment channels, eliminating financial barriers for developers worldwide. The platform supports conventional fiat payment methods such as Visa, WeChat Pay, Alipay, and UnionPay, while also offering an efficient Web3 crypto settlement network, enabling developers globally to acquire computing power with minimal friction.
The rapid advancement of large model technology is continuously expanding the boundaries of AI productivity, while the leap in underlying infrastructure determines whether the "intelligent economy" can truly take root. From architecture-level computational optimization and lifecycle-wide commercial incentives to a seamless global payment network, B.AI is fundamentally reshaping the logic of computing power distribution in the AI era with a full-stack breakthrough approach.
In this new era of "Token as Currency," B.AI remains committed to being the most robust digital foundation for all AI innovators, continuously advancing the development of supercomputing hubs and serving as the most reliable backbone behind countless industries. Here, accessible computing power is no longer just a grand vision—it is actively transforming into real productivity driving the intelligent leap of businesses worldwide.