定價
按用量付費， 無訂閱。
Anthropic、OpenAI、Google 官方原價直傳——無任何加價。國產模型僅收極薄的基礎設施溢價，所有支付通道 0 手續費。一次儲值，全網通用。
所有儲值通道手續費
Anthropic / OpenAI / Google 加價
新用戶免費額度
最低儲值金額
逐模型計價
美元 / 每百萬 token，每週根據上游官方價更新。
對話
| Model | Input | Output |
|---|---|---|
DeepSeek V3.2CN GPT-4-class general model. The most cost-effective model in the world for code & chat. | $0.1400/1M | $0.2800/1M |
Kimi K2.5CN Long-context tool-use champion. Strong agentic behaviour, free 128K context. | $0.6000/1M | $2.50/1M |
Qwen3 MaxCN 256K context. Strong on multilingual and bilingual reasoning. | $0.6000/1M | $2.40/1M |
MiniMax M2.5CN Fastest Chinese chat model — 130 tokens/sec on cold start. | $0.3000/1M | $1.20/1M |
Claude Sonnet 4.6Global Workhorse Anthropic model. 5× cheaper than Opus, ~90% the quality. | $3.00/1M | $15.00/1M |
GPT-5.5Global Latest GPT generation. Pass-through pricing. | $5.00/1M | $20.00/1M |
Gemini 3 ProGlobal 2M token context. Pass-through pricing. | $1.25/1M | $5.00/1M |
推理
| Model | Input | Output |
|---|---|---|
DeepSeek R1CN Open reasoning model rivalling o1 on math & logic, at 1/30 the price. | $0.5500/1M | $2.19/1M |
Claude Opus 4.7Global Frontier reasoning. Pass-through pricing — exactly the official rate. | $15.00/1M | $75.00/1M |
o1Global OpenAI flagship reasoning. Pass-through pricing. | $15.00/1M | $60.00/1M |
程式
| Model | Input | Output |
|---|---|---|
GLM-4.6CN Coder-tuned. Strong on Chinese function calling and tool routing. | $0.5000/1M | $1.50/1M |
Qwen3 CoderCN Coder-specialised Qwen variant. SOTA on HumanEval-CN. | $0.3000/1M | $1.50/1M |
視覺
| Model | Input | Output |
|---|---|---|
Doubao 1.6 ProCN Strong on image understanding & Chinese OCR. Mature production model. | $0.4000/1M | $1.20/1M |
GPT-4oGlobal Multimodal frontier. Pass-through pricing. | $2.50/1M | $10.00/1M |
影片
| Model | Input | Output |
|---|---|---|
Kling 1.6CN Best Chinese text-to-video. 5s/10s clips, 720p/1080p output. | $0.5000 per second of video | — |
全部支付通道，零手續費。
多數 API 聚合站收 5–6% 儲值費。我們不收。我們靠國產模型的批發差價賺錢，而不是你的錢包。
USDT (TRC-20 / ERC-20)
OpenRouter 5%
Stripe (Card)
OpenRouter 5.5% credit card
We absorb the 2.9% gateway fee
WeChat Pay / Alipay (China rails)
Not supported by OpenRouter
Bank wire + VAT invoice (China B2B)
Enterprise only on most platforms
國內對公轉帳 + 自動開 6% 增值稅電子發票
儲值後 7 個工作日內發票送達。公司抬頭、稅號、銀行帳號資訊一次儲存。
計價常見問題
Why is there a markup on Chinese models?
Upstream Chinese providers charge in CNY, demand prepaid corporate accounts, and rate-limit aggressively. We absorb that infra cost (Singapore + Shanghai egress, retry logic, multi-channel failover) and pass through to you at one transparent rate. The markup ranges 0–80% by group; check each model on the table above.
Why do you not mark up GPT / Claude / Gemini?
These vendors take direct USD payment and have global infra. We have nothing to add — we just route. The only reason to use Routify for them is to consolidate billing and keep one wallet across all models.
Do you keep my data?
No. By default, prompts and completions are not logged. Aggregate metadata (token counts, latency, model id) is retained for billing for 90 days. You can opt into per-request logging for debugging from the dashboard.
What about volume discounts?
Above $1k/month spend, you qualify for a VIP rate (5% discount across all models). Above $10k/month, contact us for enterprise terms with dedicated channels and SLA.
Can I bring my own API key (BYOK)?
Yes — coming Q3 2026. You will be able to attach your own Anthropic / OpenAI / Google key and pay only the routing fee (0% on the first 1M requests/month).
How is billing computed?
Per-request, per-token. Each response carries `X-Routify-Cost-USD` and `X-Routify-Cost-CNY` headers so your code can audit immediately. Dashboard shows every request line-item, exportable as CSV/JSON.