Loading market data...

news headline. Output JSON.

news headline. Output JSON.

. Let's draft. Lead: B.AI crossed 2 trillion cumulative tokens on August 24, capping a week of free access that pushed daily usage to a record 220 billion tokens. The campaign, which began with DeepSeek V4 Flash on August 17, has since added Tencent Hy3, DeepSeek-V4-Flash-Vision-Exp and Xiaomi MiMo-V2.5. That's good. Then section. Section 1: "A week of free models" - describe the campaign. Mention that the free access was a direct response to DeepSeek's price hikes. Also mention the record day. Section 2: "How the hybrid API is priced" - explain Official and Custom tiers, discounts, Web2/Web3 payments, rebates. Section 3: "Routing queries to the right model" - Auto mode, dynamic routing, global compute distribution hub. Section 4: "Why the free access matters" - contrarian strategy, response to price hikes. End: "B.AI has not said which model will be offered free next." We need to avoid "It's worth noting" etc. Use contractions. Let's write. I'll write the content in a natural style. Draft:

B.AI crossed 2 trillion cumulative tokens on August 24, capping a week of free access that pushed daily usage to a record 220 billion tokens. The campaign, which began with DeepSeek V4 Flash on August 17, has since added Tencent Hy3, DeepSeek-V4-Flash-Vision-Exp and Xiaomi MiMo-V2.5.

A week of free models

The free-access run started with DeepSeek V4 Flash on August 17, and that day alone saw 220 billion tokens processed — a single-day record for the platform. B.AI followed up with Tencent Hy3 on August 21, DeepSeek-V4-Flash-Vision-Exp on August 22, and Xiaomi MiMo-V2.5 on August 24, the same day the cumulative count hit 2 trillion.

The campaign was a direct response to DeepSeek's price hikes, according to the company. B.AI's move is contrarian: while competitors raise prices, it's giving away access to some of the most in-demand models.

How the hybrid API is priced

B.AI's API runs on a hybrid model with two tiers. The Official tier offers discounts of 10-40% off market rates. The Custom tier goes further, with discounts up to 90%.

The platform also supports Web2 and Web3 payment rails, and users get rebates on top of the discounted pricing. That combination is designed to pull in developers who want flexibility in how they pay.

Routing queries to the right model

B.AI's Auto mode in the Chat interface dynamically routes each query to the optimal underlying model based on intent. The company describes itself as a global compute distribution hub, shifting workloads across networks to keep costs down and performance up.

That routing is what makes the free-access campaign feasible, the company says. By spreading traffic across models and regions, B.AI can absorb the cost of free tokens while still keeping the service responsive.

Why the free access matters

The timing is deliberate. DeepSeek's price hikes gave B.AI an opening to undercut the market. Instead of matching the increases, B.AI went the other way, offering free access to models that competitors charge for.

The result is a surge in usage that's hard to ignore. Two trillion tokens in a single campaign is a scale that most platforms only dream of.

B.AI has not said which model will be offered free next.

That's about 300 words? Let's count. Actually we need 500-800. This is too short. We need to expand. We can add more details from facts. We have the facts about the campaign dates, the record, the tiers, the payment, the routing, the response. We can elaborate on each. Also we can mention the specific models and dates. We can also mention that the campaign is sitewide. We can add a sentence about the cumulative count. We can also mention that the free access is part of a broader strategy. Let's expand. We can add a paragraph about the significance of 2 trillion tokens. But we don't have context. We can say "The 2 trillion figure covers all tokens processed since the platform's launch" but we don't know that. Actually the fact says "cumulative token throughput" - so it's cumulative. We can say "The 2 trillion mark represents all tokens processed since B.AI's inception" but we don't know inception date. We can say "The cumulative count includes all tokens processed through the platform" - that's fine. We can also add a paragraph about the Web2/Web3 integration. We can explain what that means? But we shouldn't invent. We can say "The dual-rail payment system lets users pay with traditional methods or blockchain-based ones." That's an interpretation but it's reasonable. But we need to be careful not to add facts not given. The fact says "supports Web2 and Web3 dual-rail payment integrations" - we can say "supports both traditional and blockchain-based payments" but that's an interpretation. It's safe to say "supports both Web2 and Web3 payment methods" without explaining. We can also add a sentence about the Auto mode: "The Auto mode is designed to pick the best model for each query, which helps manage costs during high-traffic periods." That's an inference but reasonable. We can also add a sentence about the company's positioning: "B.AI calls itself a global compute distribution hub, a label that reflects its focus on routing compute across networks." That's fine. We can also add a sentence about the campaign's impact: "The free-access campaign has drawn developers who might otherwise have paid for these models." That's an inference but plausible. We need to avoid speculation. Let's stick to facts and reasonable paraphrasing. Let's write a longer version. I'll write: Lead: B.AI crossed 2 trillion cumulative tokens on August 24, capping a week of free access that pushed daily usage to a record 220 billion tokens. The campaign, which began with DeepSeek V4 Flash on August 17, has since added Tencent Hy3, DeepSeek-V4-Flash-Vision-Exp and Xiaomi MiMo-V2.5. Then section: "A week of free models" - expand: The free-access run started with DeepSeek V4 Flash on August 17, and that day alone saw 220 billion tokens processed — a single-day record for the platform. B.AI followed up with Tencent Hy3 on August 21, DeepSeek-V4-Flash-Vision-Exp on August 22, and Xiaomi MiMo-V2.5 on August 24, the same day the cumulative count hit 2 trillion. The campaign is sitewide, meaning every user gets free access to these models during the window. Then paragraph: The campaign was a direct response to DeepSeek's price hikes, according to the company. B.AI's move is contrarian: