Qwen-Max-0428 Release Notes

Qwen has released Qwen-Max-0428, an instruction-tuned model designed for chat services. This model represents a performance increase over the previous largest Qwen model, Qwen1.5-110B-Chat, and has achieved a top-10 ranking on the Chatbot Arena leaderboard.

Performance Benchmarks

Qwen-Max-0428 demonstrates improved capabilities over its predecessor, Qwen1.5-110B-Chat, across two primary metrics: MT-Bench and the Arena leaderboard.

Models	MT-Bench	Arena
Qwen1.5-110B-Chat	8.88	1172
Qwen-Max-0428	8.96	1186

Accessibility and Integration

Qwen-Max-0428 is available through multiple channels for testing and integration:

API Access

The model is accessible via the official DashScope API. The DashScope API is now compatible with the OpenAI API format, allowing for easier integration into existing workflows.

Interactive Demos

A public demo of the chat service is hosted on Hugging Face Spaces.

Web and App Services

The model is integrated into the Qwen web service and app, though access to these specific platforms is limited to mainland China.

Implementation Example

For developers using the OpenAI-compatible DashScope API, the following Python code demonstrates how to call the model:

from openai import OpenAI

client = OpenAI(
    api_key="$your-dashscope-api-key",
    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1"
)

completion = client.chat.completions.create(
    model="qwen-max",
    messages=[{'role': 'system', 'content': 'You are a helpful assistant.'},
              {'role': 'user', 'content': 'Tell me something about large language models.'}]
)
print(completion.choices[0].message)