Instructions to use batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8")
model = AutoModelForMultimodalLM.from_pretrained("batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8

SGLang

How to use batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8 with Docker Model Runner:
```
docker model run hf.co/batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8
```

VLLM 0.18.0 Fails to Disable Thinking Mode for Qwen3.5-35B-A3B-FP8

by huang123chuan - opened Mar 29

Discussion

huang123chuan

Mar 29

Issue: VLLM 0.18.0 Fails to Disable Thinking Mode for Qwen3.5-35B-A3B-FP8

Environment:

vllm version: 0.18.0
Model: batsclamp/Huihui-Qwen3.5-35B-A3B-Claude-4.6-Opus-abliterated-FP8

Problem Description:

When deploying the Qwen3.5-35B-A3B-FP8 model using vllm 0.18.0, the "thinking mode" cannot be disabled, even though the enable_thinking parameter is properly configured in chat_template_kwargs.

According to the official documentation, the following configuration should successfully disable thinking mode:

{
  "chat_template_kwargs": {
    "enable_thinking": false
  }
}

Expected Behavior:

When enable_thinking: false is set in chat_template_kwargs, the thinking mode should be disabled
The response should not include the thinking/thought section

Actual Behavior:

Thinking mode remains enabled despite setting enable_thinking: false
The response continues to include the thinking/thought section even after configuration

Steps to Reproduce:

Deploy Qwen3.5-35B-A3B-FP8 using vllm 0.18.0 with the following configuration:

{
  "chat_template_kwargs": {
    "enable_thinking": false
  }
}

Send a chat request to the model
Observe the response

Relevant Configuration:

{
  "chat_template_kwargs": {
    "enable_thinking": false
  }
}

Expected vs Actual:

Setting	Expected Output	Actual Output
enable_thinking: false	No thinking section	Thinking section still present

Additional Notes:

The enable_thinking: true setting appears to work correctly
The issue is specific to the enable_thinking: false configuration
This may be a bug in how vllm 0.18.0 handles the chat_template_kwargs for this particular model

huang123chuan

Mar 29

Cannot perform function call tool invocation; vllm startup includes --enable-auto-tool-choice and --tool-call-parser qwen3_coder

batsclamp

Owner Mar 29

Qwen3.5 works a bit differently here.

This will disable thinking:

curl http://localhost:8000/v1/chat/completions
-H "Content-Type: application/json"
-d '{
"model": "qwen3.5-35b",
"messages": [{"role": "user", "content": "Hello!"}],
"max_tokens": 1024,
"extra_body": {"chat_template_kwargs": {"enable_thinking": false}}
}'

This is how it looks in Cherry Studio, for example:

batsclamp

Owner Mar 29

Cannot perform function call tool invocation; vllm startup includes --enable-auto-tool-choice and --tool-call-parser qwen3_coder

I don't have those issues. Tools work fine. Pls make sure you've checked the recipe here: https://github.com/ageev/AI/blob/main/spark-recipes/batsclamp-qwen35.yaml

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment