Reasoning value
Good fit for teams comparing cost-effective Chinese reasoning models for overseas products.
Run DeepSeek chat, reasoning and coding workloads through one OpenAI-compatible gateway built for overseas developer teams.
Live · OpenAI-compatible Chat Completions
Model ID
deepseek-chatBase URL
https://thesmartoken.com/v1Input / 1M
$0.3
Output / 1M
$1.2
curl --fail-with-body -sS "https://thesmartoken.com/v1/chat/completions" \
-H "Authorization: Bearer $SMTOKEN_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-chat",
"messages": [{"role":"user","content":"Reply with exactly OK."}],
"max_tokens": 8,
"thinking": {"type":"disabled"}
}'DeepSeek global route
OpenAI-compatible model gateway
Client
Your App
SDK, backend, agent or workflow.
SmarToken
Unified API Gateway
deepseek-chat
Provider
DeepSeek Flash
DeepSeek route through SmarToken.
01
Your app
Send a standard Chat Completions request with your SmarToken key.
02
SmarToken gateway
Validate the key, model scope, daily budget and monthly budget.
03
Model route pool
Choose an enabled route by priority and weight. No automatic fallback or cross-provider failover is promised.
04
DeepSeek
Call deepseek-chat, stream the response and record usage.
Example JSON response
200 OK{
"model": "deepseek-chat",
"choices": [{
"message": {
"role": "assistant",
"content": "DeepSeek is a practical choice for coding assistants, support agents and high-volume reasoning tasks."
},
"finish_reason": "stop"
}],
"usage": { "prompt_tokens": 10, "completion_tokens": 12, "total_tokens": 22 }
}Fast Deployment
Generate a key and run the Playground in minutes.
Global Access
Use one endpoint for configured China-first routes.
Unified Billing
Track USD token usage and wallet debits together.
Easy Migration
Keep OpenAI SDK shape, change baseURL and model.
Developer Friendly
Copy-ready cURL, Python and Node.js examples.
Model fit
A focused page for evaluating this model as an API route, not just reading a catalog row.
Supplier capabilities described here do not mean every feature is supported by this gateway. Current support is text Chat Completions; tools, structured output and multimodal requests are rejected. Review the gateway contract.
Good fit for teams comparing cost-effective Chinese reasoning models for overseas products.
Useful for code generation, debugging, refactoring and software-agent prototypes.
Low displayed USD token prices help with high-volume evaluation before scaling traffic.
Route fit
Treat deepseek-chat as a route to test against a workload, not just a catalog name.
DeepSeek Flash is strongest for Reasoning, coding, research. Validate it with real prompts before making it a default route.
If your workload depends on long context or low cost, compare alternative catalog models with the same prompts.
Exact upstream availability depends on the configured provider key or aggregator channel.
DeepSeek specifics
DeepSeek's pricing model, fast release cadence and split between chat and reasoner lines change how the route should be evaluated compared with other Chinese model families.
DeepSeek's upstream platform prices cache-hit input tokens far below cache-miss tokens. Stable system prompts and repeated context lower effective cost, so check usage logs on real traffic instead of estimating from list prices alone.
deepseek-chat targets general chat and coding. DeepSeek's reasoning line trades latency for deliberate multi-step output. Send only reasoning-heavy jobs to a reasoner route instead of defaulting all traffic to it.
DeepSeek iterates quickly, and aliases, context limits and prices move between V3-era and V4-era releases. Pin the model ID you evaluated and re-run your prompt set after upstream version changes before trusting old benchmarks.
Why SmarToken
Direct accounts can work once region, billing, model IDs and credentials are settled. SmarToken is built for faster overseas evaluation and safer early production.
One API key works across the Chinese model catalog.
Model IDs, SDK samples and error semantics are written for overseas teams.
Daily, monthly and model-family limits keep experiments predictable.
Usage logs connect model, API key, token estimate, latency and wallet debit.
Code sample
curl
curl --fail-with-body -sS "https://thesmartoken.com/v1/chat/completions" \
-H "Authorization: Bearer $SMTOKEN_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-chat",
"messages": [{"role":"user","content":"Reply with exactly OK."}],
"max_tokens": 8,
"thinking": {"type":"disabled"}
}'python
import os
from openai import OpenAI
client = OpenAI(api_key=os.environ["SMTOKEN_API_KEY"],
base_url="https://thesmartoken.com/v1", max_retries=0)
response = client.chat.completions.create(
model="deepseek-chat",
messages=[{"role": "user", "content": "Reply with exactly OK."}],
max_tokens=8,
extra_body={"thinking": {"type": "disabled"}},
)
print(response.choices[0].message.content)
print(response.usage)Node.js (JavaScript)
import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.SMTOKEN_API_KEY,
baseURL: "https://thesmartoken.com/v1", maxRetries: 0,
});
const response = await client.chat.completions.create({
model: "deepseek-chat",
messages: [{ role: "user", content: "Reply with exactly OK." }],
max_tokens: 8,
thinking: { type: "disabled" },
});
console.log(response.choices[0].message.content);
console.log(response.usage);Live routes plus a few nearby catalog peers. Browse the full catalog.
| Model | API ID | Context | Input / 1M | Output / 1M |
|---|---|---|---|---|
| DeepSeek V4.1 Flash | deepseek-v4.1-flash | 1M | $0.3 | $1.2 |
| DeepSeek V4.1 Flash (Vendor Direct) | deepseek-tokenhub-flash | 1M | $0.3 | $1.2 |
| DeepSeek V4 Pro (TokenHub) | deepseek-tokenhub-pro | 1M | $1.74 | $3.48 |
| Kimi K3.1 Pro | kimi-k3.1-pro | 1M | $2 | $10 |
| Kimi K2.7 Code | kimi-k2.7-code | 256K | $0.95 | $4 |
| DeepSeek Flash | deepseek-chat | 1M | $0.3 | $1.2 |
| DeepSeek V4 Pro | deepseek-v4-pro | 1M | $1.32 | $3.96 |
| Kimi K3 | kimi-k3 | 1M | $3 | $15 |
| GLM-5.2 | glm-5.2 | 1M | $1.4 | $4.4 |
| GLM-5.3 | glm-5.3 | 1M | $1.4 | $4.4 |
| MiniMax M2.7 | minimax-m2 | 200K | $0.3 | $1.2 |
| Kimi K2 | kimi-k2 | 256K | $0.6 | $2.5 |
| Kimi K2.6 | kimi-k2.6 | 262K | $0.95 | $4 |
| Qwen Plus | qwen-plus | 128K | $1.2 | $4.8 |
| Qwen3.7 Plus | qwen3.7-plus | 1M | $0.276 | $1.101 |
SmarToken gives overseas developers one English console, one OpenAI-compatible key, Playground testing and spend controls for DeepSeek routes.
Set the SDK baseURL to https://thesmartoken.com/v1, use your SmarToken API key, and send model: deepseek-chat.
Use cases
Start with a clear workload, then compare routes in Playground before moving traffic into production.
Generate, explain and refactor code with a low-cost model route.
Answer product questions with streaming UX and predictable token spend.
Create outlines, summaries and structured reasoning notes.
Draft text summaries, labels or next steps for internal workflows.
Migration
The integration path stays familiar: same Chat Completions shape, one new baseURL and a China-first model ID.
Switch path
Keep the SDK. Change baseURL to SmarToken and switch the model ID.
Switch path
Move from broad routing into a focused Chinese-model console with budgets.
Switch path
Keep Chat Completions while adding China-first model pages and route control.
FAQ
Use the standard OpenAI SDK, set baseURL to https://thesmartoken.com/v1, authenticate with your SmarToken key, and pass model: deepseek-chat.
DeepSeek is a good candidate for reasoning, coding assistants and high-volume automation. Test it in the Playground before routing production traffic.
Use the comparison table as a starting point, then run your own prompts because coding, reasoning, latency and cost can vary by upstream route.
Yes. Use your SmarToken key, set baseURL to https://thesmartoken.com/v1, and pass the model ID shown on this page.
Input and output tokens are priced separately in USD per 1M tokens. Final usage is recorded after the provider returns a response or a stream finishes.
Create an account, open the console, generate an API key and copy the cURL, Python or Node.js example from Quickstart.
Yes. Console API keys can be limited by allowed model family plus daily and monthly USD budgets.
Start now
One API key works across the Chinese model catalog.