Instructions to use jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128 with Transformers:

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

pipe = pipeline("image-text-to-text", model="jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128", trust_remote_code=True)
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, AutoModelForCausalLM

processor = AutoProcessor.from_pretrained("jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128", trust_remote_code=True, device_map="auto")
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 jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128",
		"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/jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128

SGLang

How to use jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128 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 "jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128" \
    --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": "jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128",
		"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 "jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128" \
        --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": "jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128",
		"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 jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128 with Docker Model Runner:
```
docker model run hf.co/jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128
```

Qwen3.6-35B-A3B-DASHQ-INT4-g128

File size: 2,089 Bytes

23f15a6
 
b9e0518
23f15a6
42935da
23f15a6
 
 
 
 
42935da
23f15a6
42935da
 
23f15a6
 
42935da
 
23f15a6
e40a932
23f15a6
e40a932
 
23f15a6
e40a932
 
 
 
 
 
 
 
 
23f15a6
 
6aaec17
 
 
23f15a6
577a44a
e40a932
577a44a
 
 
 
 
 
23f15a6
 
 
 
 
 
42935da
 
 
 
23f15a6
42935da
23f15a6
42935da

---
license: apache-2.0
pipeline_tag: image-text-to-text
base_model: Qwen/Qwen3.6-35B-A3B
base_model_relation: quantized
library_name: transformers
tags:
- dashq
- quantized
- post-training-quantization
- int4
---
![DASH-Q](https://raw.githubusercontent.com/JaeminK/dashq/main/assets/dashq_banner.png)

# Qwen3.6-35B-A3B-DASHQ-INT4-g128

> **DASH-Q** — Diagonal-Aware Shrinkage for Robust PTQ.
> `INT4` · group size 128 · **20.8421 GB** (from 71.9038 GB — **3.4x smaller**)

## Usage

```python
from transformers import AutoModelForImageTextToText, AutoTokenizer

model = AutoModelForImageTextToText.from_pretrained(
    "jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128", trust_remote_code=True, device_map="cuda", dtype="auto"
)
tokenizer = AutoTokenizer.from_pretrained("jkim96/Qwen3.6-35B-A3B-DASHQ-INT4-g128")

messages = [{"role": "user", "content": "Explain 2-bit quantization in one sentence."}]
text = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)
inputs = tokenizer(text, return_tensors="pt").to(model.device)
print(tokenizer.decode(model.generate(**inputs, max_new_tokens=256)[0]))
```

`trust_remote_code=True` is required: the checkpoint ships its quantized-layer
implementation (`modeling_dashq.py`) and Triton kernels (`dashq_kernel.py`).
Without Triton, or on CPU, it falls back to dequantize-and-matmul in PyTorch.

### Requirements

| Package | Minimum | Verified with |
| --- | --- | --- |
| `torch` | 2.4 | 2.12.1+cu130 |
| `transformers` | 5.8 | 5.9.0 |
| `triton` | 3.0 (Linux; bundled with CUDA builds of PyTorch) | 3.7.1 |
| `huggingface_hub` | 1.5 (pulled in by transformers) | 1.15.0 |

## Quantization

| Field | Value |
| --- | --- |
| Base model | `Qwen/Qwen3.6-35B-A3B` |
| Precision | INT4, group size 128 |
| Scale / zero dtype | float16 |
| Calibration | wikitext2, 128 samples x 2048 |
| Size | 20.8421 GB  ·  original 71.9038 GB  ·  3.4x compression |

## Benchmarks

Full zero-shot / few-shot results for every DASH-Q checkpoint:
**[github.com/JaeminK/dashq#benchmarks](https://github.com/JaeminK/dashq#benchmarks)**