Block AI (b.ai)
Listed at https://b.ai
💰 Token Pricing
| Type | Price | Note |
|---|---|---|
| Input | 主流模型 + 0-10% 加价（如 GPT-4o $2.5→$2.75/M） | per million tokens |
| Output | 主流模型 + 0-10% 加价 | per million tokens |
🤖 Supported Models (150)
✨ Pros
- ✓One key for GPT-4o / Claude / Gemini overseas models
- ✓Direct China access, no proxy required
- ✓Developer-friendly, easy integration
- ✓Access overseas models without VPN
- ✓API Key distribution model, suitable for reselling
⚠️ Cons
- ×Platform information opaque, lacks public data
- ×Overseas models with markup may cost more
- ×Limited free credits
- ×Minimal documentation and community support
- ×Long-term stability questionable (proxy business model risk)
🎯 Best For
Developers in China who can't use VPN but need overseas models; API Key resellers
💰 Pricing & Plans
| Model | List Price (Input $ / M) | B.AI Reseller (Input $ / M) | Markup | Notes |
|---|---|---|---|---|
| GPT-4o | $2.50 | ~$2.75 | +0–10% | Most popular; markup fluctuates with rate card |
| GPT-4o-mini | $0.15 | ~$0.17 | +0–10% | Cheapest OpenAI tier on B.AI |
| Claude 3.5 Sonnet | $3.00 | ~$3.30 | +0–10% | Anthropic flagship via B.AI key |
| Claude 3.5 Haiku | $0.80 | ~$0.88 | +0–10% | Low-latency Anthropic model |
| DeepSeek-V3 | $0.27 | ~$0.30 | +0–10% | Cheapest coding-tier model |
| DeepSeek-R1 | $0.55 | ~$0.60 | +0–10% | Reasoning model; popular for agents |
| Gemini 2.5 Flash | $0.30 | ~$0.33 | +0–10% | Google fast tier |
| Gemini 2.5 Pro | $1.25 | ~$1.38 | +0–10% | Google reasoning tier |
| Mistral Large | $2.00 | ~$2.20 | +0–10% | Mistral flagship |
🔧 API & Developer Experience
- •Single API Key: One B.AI key unlocks GPT-4o, Claude, Gemini, DeepSeek, Mistral and 150+ models — no need to sign up separately with each overseas provider or wire money abroad.
- •OpenAI-Compatible Endpoint: The REST endpoint mirrors OpenAI's /v1/chat/completions schema, so existing OpenAI/Anthropic SDKs work after a base-URL and key swap. Tool/function calling and streaming are supported.
- •No-VPN China Access: B.AI's edge proxies mainland-routable HTTPS endpoints, so developers in China get OpenAI/Anthropic-grade models without setting up a VPN or overseas server.
- •Crypto-Native Billing: Top up via USDT and on-chain settlement (x402 / 8004 protocols) — useful for teams and agents that already operate on-chain and want a self-custody-friendly payment rail.
- •API Key Reseller Model: An affiliate/referral program lets developers buy in bulk and re-sell B.AI credits to downstream users, with margin kept on top of the 0–10% B.AI markup.
- •OpenClaw Integration: First-class support for the OpenClaw agent framework, including a one-click script that wires OpenClaw agents to B.AI without manual endpoint configuration.
- •Limited-Time Free Campaigns: Rotating promos (e.g. DeepSeek V4 Flash, Hy3) give free or deeply discounted access to specific models for a window — useful for evaluation before committing credits.
🌐 Web3-Native AI Reseller with Crypto Settlement
B.AI is best understood not as a model lab but as a Web3-native reseller that turns overseas LLM APIs into a single, crypto-settled key. Where OpenRouter and Cloudflare AI Gateway focus on routing and observability, B.AI's differentiator is its on-chain payment rail — top-ups in USDT, x402 and 8004 protocol settlement, and an agent wallet (BAIclaw) that lets autonomous agents self-custody funds and procure compute without a human intermediary. The LLM Service itself proxies mainstream overseas models with a 0–10% markup, packaged behind an OpenAI-compatible endpoint so existing SDKs drop in. The OpenClaw integration and invitation-rewards program turn the platform into a distribution layer as well: developers buy credits in bulk and resell them downstream, while the affiliate flow gives every signup a commission trail. For China-based teams the practical appeal is the lack of VPN requirement, but the deeper story is the convergence of AI APIs with on-chain agent commerce.
🌐 China Access & Latency
B.AI explicitly markets itself as a no-VPN China-direct route to OpenAI / Anthropic / Google overseas models. The platform exposes mainland-routable HTTPS endpoints, so most China-based ISPs reach the API directly without proxying or running overseas servers — a strong contrast to OpenAI and Anthropic which are blocked at the network layer. Edge routing is optimized for Chinese operators (China Telecom / Unicom / Mobile), so chat-completion latency is typically comparable to calling a domestic provider, often well under 200 ms for first-token time. The catch is that B.AI is fundamentally a reseller: if its upstream quota or billing channel breaks, you inherit the outage. Pricing is denominated in USD (settled via USDT), which is fine for crypto-native teams but adds a FX step for CNY-only shops. Bottom line: the easiest on-ramp to overseas models from mainland China, with the usual reseller caveats around long-term stability.