Instructions to use Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived with Transformers:

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

pipe = pipeline("text-generation", model="Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived")
model = AutoModelForCausalLM.from_pretrained("Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived

SGLang

How to use Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived 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 "Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived" \
    --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": "Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived" \
        --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": "Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived with Docker Model Runner:
```
docker model run hf.co/Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived
```

Qwen3-Next-80B-A3B-Instruct W2A16 AutoRound-derived

This experimental checkpoint was deterministically derived from the DynaExQ W4A16 AutoRound checkpoint by converting eligible expert weights to packed INT2 while preserving the mixed-precision fallbacks described in quantization_config.json. The original base model is Qwen/Qwen3-Next-80B-A3B-Instruct.

The repository includes the conversion provenance and all indexed safetensors shards. This is a research artifact; backend support for its mixed INT2 packing must be verified before deployment.

Code, conversion scripts, verification manifests, experiment results, and paper sources are available in DynaQuant. The original model license and usage restrictions continue to apply.

Downloads last month: 13

Safetensors

Model size

11B params

Tensor type

BF16

I32

F16

Model tree for Kris2017/Qwen3-Next-80B-A3B-Instruct-W2A16-AutoRound-derived

Base model

Qwen/Qwen3-Next-80B-A3B-Instruct

Quantized

Kris2017/Qwen3-Next-80B-A3B-Instruct-W4A16-AutoRound

Quantized

(1)

this model