One API key, many models: routing DeepSeek V4 and Qwen through a single endpoint

biao tang — Mon, 03 Aug 2026 08:26:12 +0000

One API key, many models

If you build on LLMs, you've probably hit this wall: DeepSeek V4 Flash is excellent for cheap reasoning, Qwen models shine on certain tasks, and you end up signing up for several providers, juggling multiple API keys, and pointing your client at different base URLs.

What if you could use one OpenAI-compatible endpoint and switch models by changing a single field?

The gateway

deeprouteapi.com is an OpenAI-compatible gateway that fronts DeepSeek V4 (Flash/Pro) and the Qwen series behind a single endpoint. Same /v1/chat/completions you already know — you just change model.

Quick start

1. Get a key

Sign up at deeprouteapi.com
Create a token in the dashboard
New accounts start with a free trial balance so you can try before paying

2. Call it with curl

curl https://deeprouteapi.com/v1/chat/completions \
  -H "Authorization: Bearer $DEEPROUTE_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-flash",
    "messages": [{"role": "user", "content": "Explain what a JWT is in one sentence."}]
  }'

Switch to a Qwen model when the task calls for it — just change the model field:

"model": "qwen-max"

3. Use it from Python

Works with the official OpenAI SDK — only the base_url and key differ:

from openai import OpenAI

client = OpenAI(
    base_url="https://deeprouteapi.com/v1",
    api_key="sk-...",  # your token
)

resp = client.chat.completions.create(
    model="deepseek-v4-flash",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(resp.choices[0].message.content)

Available models

Model	Type	Good for
`deepseek-v4-flash`	reasoning	cheap, high-volume workloads
`deepseek-v4-pro`	reasoning	harder problems, better depth
`qwen-max`	general	strong multilingual generation
`qwen-plus` / `qwen-turbo`	general	balanced / low-latency
`qwen2.5-*`	open	fine-grained size control

Billing is pay-as-you-go per token — top up when you need more, no subscription, no monthly minimum.

Why a single gateway?

One key, one endpoint for several model families
OpenAI-compatible — works with the OpenAI SDK, LangChain, and anything that speaks the chat-completions protocol
Pay as you go — only pay for what you actually use

If you're tired of managing keys across providers, give it a try — new accounts get a free trial balance.

👉 deeprouteapi.com

DEV Community: biao tang