B.AI hits 2 trillion tokens in seven days

B.AI reported that cumulative token throughput from a sitewide free-access campaign exceeded 2 trillion tokens on August 24. The company posted the milestone after a week of zero-cost access to several flagship models and upgrades to its API and payment services.

The campaign began on August 17, when B.AI opened free access to DeepSeek V4 Flash. The company recorded a single-day platform high of about 220 billion tokens that day. B.AI added Tencent Hy3 to the free tier on August 21 and DeepSeek-V4-Flash-Vision-Exp on August 22 before reporting the seven-day total on August 24.

B.AI described the activity as evidence its infrastructure handled high-concurrency, production-grade workloads. The company published the announcements in a post labeled as sponsored content and attributed platform performance to its routing and load-balancing architecture. The post did not include independent verification of the figures.

The company outlined a dual-tier API model named Official and Custom Provider. B.AI said the Official channel offers direct API connections and baseline discounts it described as between 10% and 40% off market rates. The Custom Provider option routes traffic through vetted third-party vendors including Mix, Nebula and OL Station, with up to seven discount tiers and discounts the company said can reach as much as 90% off standard pricing.

B.AI also introduced an Auto mode in its chat interface that the company says assesses user intent per prompt and routes requests to the most cost- and performance-efficient model in real time. The company framed the feature as intended to reduce unused capacity and simplify model selection for non-technical users running chat, text analytics and routine business tasks.

Additional changes described by the company include upgrades to official API services to extend price discounts across mainstream foundation models, earlier limited-time free access to MiniMax M3 and Qwen-3.8 MAX, a 10% discount on GLM 5.3, customer top-up bonuses, and payment rails combining Visa, WeChat Pay, Alipay, UnionPay and crypto settlement networks. The company said the free access windows and discount tiers aim to lower the cost barrier for enterprise AI deployments and to support developer testing of production workloads.