Community

Blog Events Webinars Tutorials Forum
  1. Blog
  2. Events
  3. Webinars
  4. Tutorials
  5. Forum
Create Account
×
Community Blog GenAI on Alibaba Cloud [Part 1]: Hello Qwen!

GenAI on Alibaba Cloud [Part 1]: Hello Qwen!

This article introduces how to build real-world AI applications using Alibaba Cloud’s Qwen model with minimal effort through a "drop-in replacement" approach.

Generative AI is shifting from "cool demos" to real-world production. But for many developers, the barrier to entry is still high. You have to learn new SDKs, worry about GPU provisioning, or deal with expensive API costs.

In this series, "GenAI on Alibaba Cloud," we are going to break down how to build powerful AI applications using the Qwen (Tongyi Qianwen, Chinese: 通义千问) model family. Qwen is consistently ranked as one of the top-performing LLMs globally, often rivaling GPT-4 in coding and reasoning benchmarks, but at a fraction of the cost.

In Episode 1, we aren't just running a "Hello World." We are going to learn the "Drop-in Replacement" technique. You will learn how to access Qwen using the standard tools you likely already know, making migration seamless.

Let's dive in.

What is Alibaba Cloud Model Studio?

Model Studio (formerly DashScope) is Alibaba Cloud’s "Model-as-a-Service" platform. Instead of buying a GPU and loading a model yourself, you simply use an API to chat with massive models running on Alibaba Cloud's infrastructure.

Why use it?

  1. Massive Context Windows: Qwen models support huge context windows (up to 1M tokens in some versions), allowing you to process entire books or codebases in one prompt.
  2. OpenAI Compatibility: As you will see below, Model Studio supports the OpenAI API format. This means your existing AI scripts can run on Alibaba Cloud just by changing the URL.
  3. Cost: Qwen-Turbo (the fast model) is incredibly affordable, and there is a generous free tier for new users.

Step 1: Get Your Credentials

Before we write code, we need a key to the engine room.

1.  Log in to the Alibaba Cloud Console.

2.  Search for "Model Studio" in the top search bar.

3.  Critical Step: If prompted to choose a region, select Singapore (or another International region close to you).

  • Why? The API endpoints differ between the China station and the International station. For global access without latency or strict firewall rules, Singapore is usually best.

4.  Navigate to Dashboard

5.  Click Key Management.

6.  Create the API Key and Copy the key (it starts with sk-). Save it somewhere safe

Step 2: The Setup

We will use Python. Instead of installing a proprietary Alibaba Cloud SDK, we will use the standard openai library.

Open your terminal and install:

pip install openai python-dotenv

Note: We install

Step 3: The Code (The "Universal" Method)

Create a file named .env in your project folder and paste your key there:

DASHSCOPE_API_KEY=sk-your_actual_key_here_no_quotes

Now, create hello_qwen.py. This script demonstrates how to route standard OpenAI commands to Alibaba Cloud's Qwen-Plus model.

import os
from openai import OpenAI
from dotenv import load_dotenv

# 1. Load environment variables
load_dotenv()

def chat_with_qwen():
    # 2. Initialize the Client
    # This is the "Magic Trick". We use the OpenAI client, but we point
    # the 'base_url' to Alibaba Cloud's International Endpoint.
    client = OpenAI(
        # Get key from .env file
        api_key=os.getenv("DASHSCOPE_API_KEY"), 
        
        # CRITICAL: This URL connects to the Singapore/Intl gateway.
        # If you are in China, you might use 'dashscope.aliyuncs.com' instead.
        base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
    )

    print("Connecting to Qwen...")

    try:
        # 3. Send the Prompt
        # We are using 'qwen-plus', a great balance of speed and intelligence.
        completion = client.chat.completions.create(
            model="qwen-plus",  
            messages=[
                {'role': 'system', 'content': 'You are a helpful expert on Cloud Computing.'},
                {'role': 'user', 'content': 'Explain the concept of "Serverless" to a 10-year-old.'}
            ]
        )
        
        # 4. Print the Response
        print("\n--- Qwen says: ---")
        print(completion.choices[0].message.content)
        print("------------------")
        
    except Exception as e:
        print(f"Error: {e}")
        print("Tip: Check if your API Key matches the region in your base_url.")

if __name__ == '__main__':
chat_with_qwen()

Step 4: Run and Verify

Run the script in your terminal:

python hello_qwen.py

You should see an output similar to:

Deep Dive: Which Model Should You Use?

In the code above, we used model="qwen-plus". However, Alibaba Cloud offers a few flavors. Here is how to choose:

Model Name Best For... Cost Profile
qwen-turbo Simple chatbots, translation, summarizing short text. Lowest (Very fast)
qwen-plus General purpose assistant, content creation, reasoning. Balanced (Standard choice)
qwen-max Complex coding, math, legal analysis, or nuance. Higher (Best performance)
qwen-vl-plus Analyzing Images (Visual Language). Varies

Pro Tip: Start developing with qwen-turbo to save money/credits, then switch to plus or max for production if you need higher quality answers.

Troubleshooting Corner

If you ran the code and got an error, check these common pitfalls:

Error:

● Cause: Your API Key region and your base_url do not match.

● Fix: If you created your key on the International site (alibabacloud.com), you must use https://dashscope-intl.aliyuncs.com/compatible-mode/v1. If you used the wrong URL, the system thinks your key doesn't exist.

Error:

● Cause: Usually an issue with the message format.

● Fix: Ensure your messages list has the correct structure: {'role': 'user', 'content': '...'}.

What's Next?

Congratulations! You have successfully integrated one of the world's most powerful LLMs into your Python application.

Since we set this up using the OpenAI-compatible client, you have unlocked a massive ecosystem. Any tool, library, or framework (like LangChain or AutoGen) that works with OpenAI will now work with Alibaba Cloud Qwen.

In Episode 2, we will put this to work. We will build a RAG (Retrieval Augmented Generation) system to make Qwen answer questions based on your own private PDF documents.

See you in the next episode!

Disclaimer: The views expressed herein are for reference only and don't necessarily represent the official views of Alibaba Cloud.

AI Artificial Intelligence Tutorials Generative AI GenAI LLMs Qwen Alibaba Cloud Model Studio Model Studio Model-as-a-Service DashScope
0 1 0
Share on

Read previous post:

Building Serverless Applications with Alibaba Cloud Function Compute

Read next post:

GenAI on Alibaba Cloud [Part 2]: Chat with Your PDF (Building a RAG System)

Farah Abdou

14 posts | 0 followers

You may also like

Comments

Farah Abdou

14 posts | 0 followers

Related Products

More Posts by Farah Abdou

See All