Could you please show me a deployment example for a qwen3 model with hybrid thinking, for example, Qwen3-30B-A3B, and example of disabling thinking

To deploy Qwen3-30B-A3B with hybrid thinking (reasoning enabled), use the following command, specifying the reasoning parser as qwen3: vllm serve Qwen/Qwen3-30B-A3B --enable-reasoning --reasoning-parser qwen3 To disable thinking (reasoning) for a specific request, pass chat_template_kwargs={“enabl…

--enable-reasoning is outdated.

Correct, the --enable-reasoning flag is deprecated and has been removed as of vLLM v0.10.0. You should use only --reasoning-parser (e.g., --reasoning-parser qwen3) to enable reasoning mode for supported models. When --reasoning-parser is specified, reasoning is automatically enabled—no need for --en…

Do I still need to set chat_template_kwargs to disable thinking?

Yes, to disable thinking (reasoning) for Qwen3 and similar models, you still need to set chat_template_kwargs={“enable_thinking”: False} in your API request. This is the recommended way to turn off reasoning on a per-request basis, even when using the latest vLLM and reasoning parser. See reasoning …

If you set chat_template_kwargs={“enable_thinking”: False} but still receive reasoning content, ensure you are using the correct reasoning parser (qwen3 for Qwen3 models) and the latest vLLM version. Some older versions or incorrect parser settings may ignore the disable flag, causing reasoning cont…

show me code examples to set reasoning config for reasoning models in vllm

To set a thinking budget (limit the number of reasoning tokens) for Qwen3 models in vLLM, pass the thinking_token_budget parameter in your API request. For the OpenAI-compatible API, include it in extra_body, e.g.: response = client.chat.completions.create( model="Qwen/Qwen3-30B-A3B", messa…

@hustxiayang while serving you can disable the reasoning with by setting“enable_thinking" as false vllm serve Qwen/Qwen3-8B –reasoning-parser qwen3 –default-chat-template-kwargs ‘{“enable_thinking”: false}’

Deployment example for a qwen3 model with hybrid thinking

Features Reasoning Outputs

hustxiayang August 26, 2025, 6:05pm 7

I set it, but still got reasoning process?

Topic		Replies	Views
Disabling reasoning of Qwen3-VL-8B-Thinking per request Model Support	1	3776	October 29, 2025
How to get thinking content in qwen3.5 thinking General	1	1012	March 19, 2026
Thinking Token limit setting General	11	887	February 26, 2026
how to disable thinking for different model? General	1	4212	September 2, 2025
Qwen3.5-27b-fp8没有think General	3	392	March 30, 2026

Deployment example for a qwen3 model with hybrid thinking

Related topics