One key for frontier and Chinese models, without the markup guesswork
Route GPT, Claude, Gemini, Kimi, GLM, DeepSeek and MiniMax through a single OpenAI-compatible endpoint. Transparent per-model multipliers, prepaid credits that never expire, and a Messages-compatible endpoint for Claude Code.
How this gateway compares to OpenRouter
Both give you many models behind one API. The differences are in billing, the model mix, and how Claude Code connects.
| This gateway | OpenRouter | |
|---|---|---|
| API surface | OpenAI Chat Completions, Responses, and Anthropic Messages compatible | OpenAI-compatible routing across many upstreams |
| Billing model | Prepaid credits in CNY, fixed per-model multiplier off a 0.4 CNY / 1M base | Prepaid balance in USD, upstream price plus a routing fee |
| Credit expiry | Credits do not expire or reset | Balance held in account |
| Chinese model access | Kimi, Qwen, GLM, DeepSeek, MiniMax, Doubao, MiMo, StepFun on the same key | Depends on which upstreams are connected |
| Claude Code support | Native Messages endpoint — point ANTHROPIC_BASE_URL and go | Via OpenAI-compatible shim |
| Self-serve purchase | Buy online, key delivered by email instantly | Self-serve top-up |
Who this is for
A good fit if you
- Want one key that spans both frontier (GPT, Claude, Gemini) and Chinese models
- Prefer prepaid credits that don't expire over a metered monthly bill
- Run Claude Code or an Anthropic-SDK app and want a drop-in Messages endpoint
- Want a fixed, published multiplier per model instead of fluctuating upstream prices
Not the right fit if you
- Need to bring your own upstream provider keys and only pay pass-through
- Require USD invoicing or a Western tax entity for procurement
- Depend on a specific model that this gateway does not route
- Need an SLA-backed enterprise contract today
Honest limitations
What to know before you switch, stated plainly.
- Pricing and prepaid credits are denominated in CNY. There is no USD billing surface today.
- The model catalog is curated, not unlimited — if you need an obscure upstream, check the pricing page first.
- Rate limits are per API key. High-concurrency workloads should use exponential backoff on HTTP 429.
- We do not publish latency or uptime SLAs we have not measured. Test your own workload before committing volume.
The numbers, from our own pricing
Every figure below is the live, published rate — not an estimate. Multipliers apply to a 0.4 CNY per 1M token base.
We deliberately omit latency and uptime claims here because they depend on your region and workload. Run your own test against the docs endpoint before moving production traffic.
OpenRouter alternative — FAQ
Is this a drop-in replacement for OpenRouter?
For most OpenAI-compatible calls, yes — point your base URL at the gateway and swap the model name. Claude Code and Anthropic-SDK apps additionally get a native Messages-compatible endpoint.
How is pricing different from OpenRouter?
Each model has a fixed published multiplier on a 0.4 CNY per 1M token base, paid from prepaid credits that don't expire. OpenRouter bills a USD balance at upstream price plus a routing fee.
Which models can I reach on one key?
GPT, Claude, Gemini, Kimi, Qwen, GLM, DeepSeek, MiniMax, Doubao, MiMo and StepFun families. See the pricing page for the current list and per-model multipliers.
Do my credits expire?
No. Purchased credits do not expire or reset, and you can top up an existing key at any time.
Try one key across every model
Buy a key, point your base URL here, and keep the credits as long as you need.