Quick Start: Your First API Call to Qwen in 5 Minutes

Qwen is Alibaba's flagship open-weight model family, and the current general-purpose release is Qwen 3.7, exposed on TokenPAPA as qwen3.7-plus. It is strong at code generation, structured output, and Chinese-language work, and it sits in the cheap tier of the market — which makes it a natural second model next to DeepSeek.

The barrier is rarely the model. It is the account layer. Alibaba Cloud Model Studio expects a verified Alibaba Cloud account and, for most overseas developers, a domestic payment method before you can call anything.

Qwen API quick start in one line: install the OpenAI SDK, set base_url to https://tokenpapa.ai/v1, set model to qwen3.7-plus, and send a chat completion request. That is the entire integration — about five minutes from zero to a response.

This is the shortest path, with no Alibaba Cloud account and no Chinese phone number.

---

What you need

Requirement	Detail
Python	3.8 or newer (or Node 18+ for the JavaScript version below)
Package	openai — the standard OpenAI SDK
Account	tokenpapa.ai — email, Google, or GitHub
Payment	International card, Apple Pay, or Google Pay (minimum top-up $10)
Base URL	https://tokenpapa.ai/v1
Model ID	qwen3.7-plus

Nothing Alibaba-specific is on that list. The Qwen API is OpenAI-compatible on TokenPAPA, so the client you already use for OpenAI or DeepSeek works unchanged.

---

Step 1 — Install the SDK (30 seconds)

pip install --upgrade openai

That is the only dependency. If you already have openai installed for another provider, skip this step entirely.

Step 2 — Get a Qwen API key (about 2 minutes)

1. Sign up at tokenpapa.ai using an email address, or use Google / GitHub one-click login (a first-time OAuth login creates the account automatically).
2. Open the console and generate an API key.
3. Store it as an environment variable — never hard-code it in a repository.

export TOKENPAPA_API_KEY="your-tokenpapa-key"

No Chinese phone number, no SMS verification code, no Alibaba Cloud account, and no domestic bank card are involved at any point.

Step 3 — Send your first request (30 seconds)

from openai import OpenAI

client = OpenAI(
    api_key=os.environ["TOKENPAPA_API_KEY"],
    base_url="https://tokenpapa.ai/v1"
)

response = client.chat.completions.create(
    model="qwen3.7-plus",              # Qwen 3.7 — 256K context
    messages=[
        {"role": "system", "content": "You are a concise technical assistant."},
        {"role": "user", "content": "Explain what a KV cache is in two sentences."}
    ],
    max_tokens=300                     # always cap output: output tokens cost more than input
)

print(response.choices[0].message.content)
print("tokens:", response.usage.total_tokens)

Run it and you should see the answer followed by a token count. If you get a 401, re-copy the key — a trailing newline in the environment variable is the usual culprit.

Key takeaway: The Qwen API on TokenPAPA is a drop-in replacement for the OpenAI API. Only three values change — base_url, api_key, and model. Every other line of your existing code, including retries, streaming, and tool calling, stays exactly as it is.

Step 4 — Parse the response properly

The response object mirrors OpenAI's shape, so you can read structured fields instead of the raw string:

choice = response.choices[0]

print(choice.message.content)   # the model's answer
print(choice.finish_reason)     # "stop" = completed, "length" = hit max_tokens
print(response.usage.prompt_tokens, response.usage.completion_tokens)

Two practical notes:

• finish_reason == "length" means the answer was cut off by your max_tokens. Raise the cap if the task needs longer output, and remember that output tokens are billed at roughly 3x the input rate on Qwen 3.7.
• Log usage from day one. It is the only reliable way to attribute cost per feature once you have more than one caller.

---

Bonus: streaming and JavaScript

Streaming is a one-flag change:

stream = client.chat.completions.create(
    model="qwen3.7-plus",
    messages=[{"role": "user", "content": "Write a haiku about rate limits."}],
    stream=True,
    max_tokens=200
)

for chunk in stream:
    delta = chunk.choices[0].delta.content
    if delta:
        print(delta, end="", flush=True)

The same endpoint works from Node:

const client = new OpenAI({
  apiKey: process.env.TOKENPAPA_API_KEY,
  baseURL: 'https://tokenpapa.ai/v1',
});

const res = await client.chat.completions.create({
  model: 'qwen3.7-plus',
  messages: [{ role: 'user', content: 'Hello Qwen.' }],
  max_tokens: 200,
});

console.log(res.choices[0].message.content);

---

Direct Alibaba Cloud vs TokenPAPA

Both routes reach the same model. The difference is entirely on the account and operations side.

Dimension	Alibaba Cloud Model Studio (direct)	TokenPAPA
Account requirement	Verified Alibaba Cloud account	Email, Google, or GitHub
Phone / identity check	Domestic identity and payment expected	None
Payment method	Domestic rails in most regions	International cards, Apple Pay, Google Pay
Minimum top-up	Varies by channel	$10
Model coverage	Qwen family	Qwen, DeepSeek, Kimi, GLM, MiniMax, GPT, Claude, Gemini
Model switching	Separate integrations per vendor	One-line model= change
API format	OpenAI-compatible (region-dependent endpoints)	OpenAI-compatible, single endpoint
Setup time	Account verification can take hours to days	About 5 minutes

When a gateway is the right call: if you are outside mainland China, need to bill on an international card, or want Qwen alongside non-Alibaba models in the same codebase, a gateway removes the account friction entirely and keeps one integration surface.

The honest caveat: direct access can be the better choice if you are inside China, already have enterprise Alibaba Cloud billing, or need Alibaba-specific features such as private model deployment or dedicated throughput. Gateway routing adds a hop, and for teams with an existing enterprise agreement that hop has no upside.

---

Qwen models you can reach with the same key

Model ID	Good at	Context
qwen3.7-plus	General flagship — chat, coding, agents, structured output	256K
qwen3.7-max	Larger Qwen 3.7 tier for heavier reasoning	see pricing page
qwen3.8-flash	Fast, low-cost Qwen tier for high-volume calls	see pricing page

For a cost baseline, Qwen 3.7 (qwen3.7-plus) is listed at $0.20 per 1M input and $0.60 per 1M output tokens, while DeepSeek V4 Flash is listed at $0.14 / $0.42 with a 128K window. Rates change — confirm current numbers on the pricing page before you build a budget on them.

---

Common errors and fixes

Error	Meaning	Fix
401 Invalid token	Key is wrong, revoked, or has whitespace	Re-copy the key; check for a trailing newline in the env var
402 Insufficient balance	Balance exhausted	Top up in the console (minimum $10)
404 model not found	Model ID does not exist	Use a live ID such as qwen3.7-plus
429 Too many requests	Rate limit hit	Add exponential backoff with jitter
400 context length exceeded	Prompt plus max_tokens exceeds the window	Trim history or move to a longer-context model

The 404 case is the most common first-day mistake, because marketing names and API model IDs rarely match. The model people call "Qwen 3.7" is called with the ID qwen3.7-plus — always confirm the exact ID before shipping.

---

FAQ

Q: What is the fastest way to get a Qwen API key?
A: Sign up at tokenpapa.ai with an email address or Google/GitHub one-click login, then generate a key in the console. No Chinese phone number, SMS code, or Alibaba Cloud account is required — the step usually takes under two minutes.

Q: Which Python SDK do I need for the Qwen API?
A: The standard openai SDK. The endpoint at https://tokenpapa.ai/v1 is OpenAI-compatible, so you set base_url and pass a Qwen model ID such as qwen3.7-plus. No Alibaba-specific client is needed.

Q: How much does the Qwen API cost on TokenPAPA?
A: Qwen 3.7 (qwen3.7-plus) is listed at $0.20 per 1M input and $0.60 per 1M output tokens with a 256K context window — about 1% of a frontier model like GPT-5.6 Sol on input price. Confirm current rates at tokenpapa.ai/pricing before budgeting.

Q: Can I call Qwen and DeepSeek from the same API key?
A: Yes. One TokenPAPA key reaches the Qwen, DeepSeek, Kimi, GLM, MiniMax, GPT, Claude, and Gemini families through the same endpoint. Switching is a one-line change to the model parameter.

---

Get Started

1. Sign up at tokenpapa.ai — email, Google, or GitHub. No Chinese phone number.
2. Create an API key in the console.
3. Point the OpenAI SDK at https://tokenpapa.ai/v1 and call qwen3.7-plus.

from openai import OpenAI

client = OpenAI(api_key="your-tokenpapa-key", base_url="https://tokenpapa.ai/v1")

print(client.chat.completions.create(
    model="qwen3.7-plus",
    messages=[{"role": "user", "content": "Hello Qwen."}],
    max_tokens=100
).choices[0].message.content)

Five minutes of setup, one key, and Qwen 3.7 is reachable from anywhere — with DeepSeek, Kimi, GLM, GPT, Claude, and Gemini on the same endpoint whenever you want to compare.

---

Originally published at https://doc.tokenpapa.ai/en/docs/blog/qwen-api-quick-start-5-minutes.