Instructions to use QuantTrio/Qwen3.6-35B-A3B-AWQ with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use QuantTrio/Qwen3.6-35B-A3B-AWQ with Transformers:

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

pipe = pipeline("image-text-to-text", model="QuantTrio/Qwen3.6-35B-A3B-AWQ")
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, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("QuantTrio/Qwen3.6-35B-A3B-AWQ")
model = AutoModelForImageTextToText.from_pretrained("QuantTrio/Qwen3.6-35B-A3B-AWQ")
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

vLLM

How to use QuantTrio/Qwen3.6-35B-A3B-AWQ with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "QuantTrio/Qwen3.6-35B-A3B-AWQ"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "QuantTrio/Qwen3.6-35B-A3B-AWQ",
		"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/QuantTrio/Qwen3.6-35B-A3B-AWQ

SGLang

How to use QuantTrio/Qwen3.6-35B-A3B-AWQ 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 "QuantTrio/Qwen3.6-35B-A3B-AWQ" \
    --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": "QuantTrio/Qwen3.6-35B-A3B-AWQ",
		"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 "QuantTrio/Qwen3.6-35B-A3B-AWQ" \
        --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": "QuantTrio/Qwen3.6-35B-A3B-AWQ",
		"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 QuantTrio/Qwen3.6-35B-A3B-AWQ with Docker Model Runner:
```
docker model run hf.co/QuantTrio/Qwen3.6-35B-A3B-AWQ
```

PPLX or KLD, or other benchmark

by HenkTenk - opened Apr 20

Discussion

HenkTenk

Apr 20

Hey, could you maybe post KLD or PPLX or other benchmarks to compare the quant to the original checkpoint?

For GGUFs there is alot of information available but for AWQs benchmark info is scarce.
https://www.reddit.com/r/LocalLLaMA/comments/1so5nrl/qwen36_gguf_benchmarks/

I also see there is a PR made to measure it in VLLM, but isn't merged yet:
https://www.reddit.com/r/LocalLLaMA/comments/1rkmvo4/i_added_ppl_and_kld_to_vllm_review_rfc_and_pr_and/

For SGlang i found this, quite involved, method:
https://github.com/voipmonitor/rtx6kpro/blob/master/benchmarks/kld-evaluation.md

tclf90

QuantTrio org Apr 21

•

edited Apr 21

Thanks for the suggestion and for sharing these links.

We do think benchmarks like KLD / PPLX are meaningful. For now though, we’re trying to stay a bit cautious. Since many of these open weight models come from commercial companies, we’d prefer not to publish systematic benchmark numbers ourselves, just to avoid possible legal/compliance issues.

Really appreciate the community helping out with benchmarking and feedback though — that’s super valuable for future quantization work.

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