vllm

vllm copied to clipboard

Your current environment

VLLM version: 0.85

🐛 Describe the bug

The command to serve Qwen3-32b:

VLLM_USE_V1=0 vllm serve Qwen/Qwen3-32B --served-model-name qwen3-32b -tp 4 --trust-remote-code --enable-reasoning --reasoning-parser deepseek_r1

Here I use VLLM_USE_V1=0 to enable guided output.

The query command:

curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
        "model": "qwen3-32b",
        "messages": [
          {"role": "system", "content": "You are a helpful assistant."},
          {"role": "user", "content": "你好"}
        ],
        "chat_template_kwargs": {"enable_thinking": false}
      }'

Here, I want to disable thinking temporarily.

And I got:

...
  "choices": [
    {
      "finish_reason": "stop",
      "index": 0,
      "message": {
        "content": null,
        "role": "assistant",
        "tool_calls": null,
        "function_call": null,
        "refusal": null,
        "reasoning_content": "你好！有什么我可以帮你的吗？"
      }
    }
  ],
...

I think the reply should be in the content field instead of the reasoning_content field.

Before submitting a new issue...

• [x] Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.