Question 1

How is billing calculated — by token or by request?

Accepted Answer

You're billed by actual token usage, with input and output tokens metered separately at each model's published rate. No prepaid holds, no monthly subscription — only pay for what you use. Per-request token consumption and balance changes are shown in real time on the usage page.

Question 2

What's the minimum top-up? Is there a minimum spend?

Accepted Answer

Minimum top-up is $10 USD, and credits never expire. No monthly fee, no minimum-spend requirement — once you verify your email at signup we credit a free trial allowance you can use immediately.

Question 3

Which AI models are supported?

Accepted Answer

We've integrated Anthropic Claude (Opus / Sonnet / Haiku — full lineup), OpenAI GPT-5 series, Google Gemini 2.5, DeepSeek, Qwen, Mistral and other mainstream models. New models are typically available within 1-7 days of release — no key migration needed.

Question 4

How does pricing compare to the official APIs?

Accepted Answer

Mainstream models are priced 10%-30% below direct official rates, and a single account covers every model — no need to register separately with OpenAI, Anthropic, or Google Cloud. Enterprise volume customers can negotiate further discounts.

Question 5

Is the OpenAI format supported? Do I need to change existing code?

Accepted Answer

Fully compatible with the OpenAI format (/v1/chat/completions, /v1/messages) and the native Anthropic Messages API. Existing code built on the OpenAI SDK or Claude SDK only needs to swap base_url and api_key — nothing else has to change.

Question 6

Can enterprises get invoices? Do you accept business bank transfers?

Accepted Answer

We issue both general and special VAT invoices, delivered 1-3 business days after a corporate bank transfer is credited. Enterprise customers can request team accounts with shared credit pool, individual API keys, and per-member usage analytics.

Question 7

Are Claude / GPT routes stable? What about latency?

Accepted Answer

We operate accelerated routing through partner networks with backbone optimizations for mainstream AI providers, keeping average latency under 200ms. Enterprise plans can apply for dedicated lines with a 99.9% SLA.

Question 8

Do coding assistants (Cursor / Cline / Claude Code) work?

Accepted Answer

Fully supported. In Cursor, set the OpenAI API Base to our endpoint to use GPT-5; for Claude Code, override the ANTHROPIC_BASE_URL and ANTHROPIC_API_KEY environment variables. Full setup details are in the developer docs.

Question 9

Will my data be used for training? How is privacy protected?

Accepted Answer

All requests are forwarded as relay traffic to the upstream provider — we never store your prompt content for training. Request logs retain only metadata (token counts, latency, status codes) for billing and troubleshooting, and are auto-purged after 30 days.

Question 10

How do I get technical support?

Accepted Answer

We offer 7×12 live chat plus a technical support group. Enterprise customers can be assigned a dedicated technical contact with average response under 10 minutes. The admin console also ships with built-in error-code explanations and self-service troubleshooting tools.

AI Model Directory

FAQ

One account, every model