Hey everyone,

​I'm running into a weird issue and hoping someone here might have a fix or some troubleshooting ideas. I'm currently trying to run the new Gemma 4 31b-it model using vLLM (v0.20.0-cu130) deployed via Helm chart (https://github.com/vllm-project/vllm/tree/main/examples/online_serving/chart-helm).

For context, this is the command I used for running vLLM:

```

command: ["vllm", "serve", "/data", "--served-model-name", "google/gemma-4-31b-it","--safetensors-load-strategy", "lazy", "--dtype", "bfloat16", "--max-model-len", "4096", "--gpu-memory-utilization", "0.8", "--host", "0.0.0.0", "--port", "8000", "--chat-template", "/data/chat_template.jinja", "--reasoning-parser", "gemma4"]

```

​When I try to send a simple message to the model using the following script:

```

from openai import OpenAI

client = OpenAI(

base_url="http://localhost:8000/v1",

api_key="",

)

response = client.chat.completions.create(

model="google/gemma-4-31b-it",

messages=[

{"role": "user", "content": "hello how are you?"}

]

)

print(response.choices[0].message.content)

```

Instead of a normal response, I keep getting this strange, repetitive output:

thinking nvarchar(max) nvarchar(max) nvarchar(max)...

Has anyone experienced this specific issue with this model or vLLM version? Any pointers on what might be causing it or how to fix my configuration would be hugely appreciated!

Thanks in advance.