Buy a Chinese LLM API key from outside China
Kimi, Qwen, GLM, DeepSeek, MiniMax, Doubao, MiMo and StepFun behind a single OpenAI-compatible endpoint. No China payment method, no per-provider signup, no separate SDKs.
Compare token costs
CNY per million tokens. This estimates metered usage, not the cash price of a prepaid package.
Make a test API call qwen3.8-max · cURL
Set GATEWAY_API_KEY in your terminal. Running this request uses credits. The response reports input and output tokens in usage.prompt_tokens and usage.completion_tokens.
curl -sS 'https://api.llm-token.cn/v1/chat/completions' \
-H "Authorization: Bearer ${GATEWAY_API_KEY:?Set GATEWAY_API_KEY}" \
-H "Content-Type: application/json" \
--data-binary @- <<'JSON'
{
"model": "qwen3.8-max",
"messages": [
{
"role": "user",
"content": "Reply with OK."
}
]
}
JSONOne gateway vs. signing up with each Chinese provider
You can integrate each Chinese model vendor directly, or reach them all through one compatible endpoint. Here is the difference for an overseas developer.
| This gateway | Direct per-provider | |
|---|---|---|
| Accounts to manage | One key for all supported Chinese models | One account and key per provider |
| API format | OpenAI Chat Completions / Responses compatible | Varies by provider, sometimes China-only SDKs |
| Payment | Prepaid API credits in USD | Often requires a domestic China payment method |
| Overseas access | Single overseas-reachable endpoint | Some endpoints are geofenced or domestic-only |
| Switching models | Change the model string, keep the code | Re-integrate per provider |
| Direct vendor features | Common surface; vendor-specific extras may not pass through | Full vendor-specific feature set |
Who this is for
A good fit if you
- Are outside mainland China and want to call Kimi, Qwen, GLM, or DeepSeek
- Don't want to manage a separate account, key, and SDK per provider
- Can't easily get a domestic Chinese payment method
- Want OpenAI-compatible calls so your existing client just works
Not the right fit if you
- Need a provider-specific feature that only the vendor's native API exposes
- Require a contract signed directly with the model vendor
- Depend on a Chinese model this gateway does not route
Honest limitations
What an aggregating gateway can and cannot do.
- Vendor-specific parameters or beta features may not pass through a compatible surface.
- Pay in USD for prepaid API credits. Model rates are listed in CNY. Usage depends on the model and its input, output and cache rates.
- The catalog is curated — confirm your target model is on the pricing page.
- We do not publish overseas latency numbers we have not measured. Latency depends on your region; test first.
MiMo-V2.6 Pro/Flash API: Pricing, Benchmarks & CodeMidas
Choose MiMo-V2.6 Pro or Flash with the AA 46-point context, $3.47M Live RL disclosure, CodeMidas method, official cached API prices, and platform multipliers.
Chinese LLM API — FAQ
Can I call Chinese models from outside China?
Yes. The gateway exposes a single overseas-reachable, OpenAI-compatible endpoint that routes to Kimi, Qwen, GLM, DeepSeek and more, so you don't depend on geofenced domestic endpoints.
Do I need a Chinese phone number or payment method?
Use international checkout to buy credits in USD with one account for the gateway.
Will my OpenAI SDK code work?
Yes for Chat Completions and Responses-style calls. Set the base URL to https://api.llm-token.cn/v1 and change the model string to the Chinese model you want.
Which Chinese models are supported?
Kimi, Qwen, GLM, DeepSeek, MiniMax, Doubao, MiMo and StepFun families today. Check the pricing page for the current list and per-model multipliers.
Get Chinese model API access today
Buy one key, point your OpenAI client at the gateway, and ask Telegram support for the fastest route if you are outside China.