China-first catalog
Pages, examples and pricing focus on Chinese model families instead of a generic global list.
OpenRouter migration
A focused OpenRouter alternative for Chinese frontier AI models
SmarToken is for teams whose OpenRouter usage is mostly Chinese model traffic. Keep the OpenAI SDK shape, move credentials to SmarToken and compare DeepSeek, Qwen, Kimi, GLM, Doubao, Hunyuan and MiniMax through one gateway built around those routes.
Low-friction migration
Common OpenRouter-style aliases are normalized while you move to documented SmarToken IDs.
Production controls
Budget-limited keys, usage logs and prepaid balance make route experiments easier to contain.
Code change
Change the base URL, key and model ID
The request shape remains familiar. For production code, use the SmarToken model ID from the catalog so logs, pricing and support tickets all refer to the same identifier.
const client = new OpenAI({
apiKey: process.env.OPENROUTER_API_KEY,
baseURL: "https://openrouter.ai/api/v1",
});
await client.chat.completions.create({
model: "deepseek/deepseek-chat",
messages,
});const client = new OpenAI({
apiKey: process.env.SMTOKEN_API_KEY,
baseURL: "https://thesmartoken.com/v1",
});
await client.chat.completions.create({
model: "deepseek-chat",
messages,
});Migration checklist
Move traffic in small, measurable steps
Treat migration as an evaluation run. If a route is cheaper but needs twice as many retries or edits, the real cost may not be lower.
- 01Create one SmarToken API key for staging and keep it server-side.
- 02Change the SDK baseURL to the SmarToken /v1 base URL.
- 03Map OpenRouter-style model names to SmarToken model IDs.
- 04Run a small prompt set that represents real workloads, not only a hello-world request.
- 05Compare quality, latency, token cost and retry behavior before moving production traffic.
- 06Create separate keys or budgets for batch jobs, user-facing features and internal tools.
Alias map
Common OpenRouter-style IDs and SmarToken routes
| SmarToken model | Provider | Compatible aliases | Best for |
|---|---|---|---|
| deepseek-chat | DeepSeek | deepseek/deepseek-chat | coding copilots, support agents |
| kimi-k2 | Moonshot AI | moonshot/kimi-k2 | research assistants, codebase analysis |
| qwen-plus | Alibaba Cloud | qwen/qwen-plus, alibaba/qwen-plus | multilingual chat, structured extraction |
| glm-4-flash | Z.ai | z-ai/glm-4-flash, thudm/glm-4-flash | internal tools, knowledge assistants |
| doubao-seed | ByteDance | bytedance/doubao-seed | consumer chat, content workflows |
| ernie-4-turbo | Baidu | baidu/ernie-4-turbo | Chinese knowledge QA, enterprise search |
| hunyuan-turbos | Tencent Cloud | tencent/hunyuan-turbos | enterprise assistants, code generation |
| minimax-m2 | MiniMax | minimax/minimax-m2 | coding agents, tool-use workflows |
| step-3 | StepFun | stepfun/step-3 | coding tools, agent platforms |
| baichuan4-turbo | Baichuan AI | baichuan/baichuan4-turbo | domain assistants, Chinese knowledge QA |
| spark-x1 | iFLYTEK | iflytek/spark-x1 | education assistants, reasoning QA |
| sensenova-v6 | SenseTime | sensetime/sensenova-v6 | multimodal assistants, visual QA |
Positioning
When SmarToken is a better fit
SmarToken is not the right answer for every routing strategy. It is strongest when China-model access, documentation and cost tracking are the central requirement.
| Dimension | OpenRouter-style router | SmarToken |
|---|---|---|
| Catalog scope | Broad global model router | Focused Chinese frontier model gateway |
| Best fit | Teams exploring many global providers | Teams building with DeepSeek, Qwen, Kimi, GLM, Doubao and similar routes |
| Migration effort | Existing OpenRouter aliases | OpenAI SDK shape plus model alias normalization for common China-model IDs |
| Spend controls | Provider and account-level controls vary | Console API keys, wallet balance and usage history in one place |
| Human-readable docs | Broad provider documentation | Model pages, pricing and tutorials focused on China-model workloads |
OpenRouter migration FAQ
Is SmarToken a full OpenRouter replacement?
No. SmarToken is narrower by design. It focuses on OpenAI-compatible access to Chinese model families such as DeepSeek, Qwen, Kimi, GLM, Doubao, Hunyuan and MiniMax instead of trying to mirror every global model.
Can I keep using the OpenAI SDK?
Yes. Keep the OpenAI SDK, set baseURL to https://thesmartoken.com/v1, use a SmarToken API key and choose a supported SmarToken model ID or compatible alias.
Should production apps use OpenRouter-style aliases forever?
Aliases are useful for migration testing. For long-term production code, prefer the documented SmarToken model ID shown on the model catalog and model detail pages.
What should I test before moving traffic?
Compare answer quality, latency, cost per accepted output, streaming behavior, error handling and how each route behaves on your own prompts.