Instructions to use cyankiwi/Qwen3-Coder-Next-AWQ-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use cyankiwi/Qwen3-Coder-Next-AWQ-4bit with Transformers:

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

pipe = pipeline("text-generation", model="cyankiwi/Qwen3-Coder-Next-AWQ-4bit")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("cyankiwi/Qwen3-Coder-Next-AWQ-4bit")
model = AutoModelForCausalLM.from_pretrained("cyankiwi/Qwen3-Coder-Next-AWQ-4bit")
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

vLLM

How to use cyankiwi/Qwen3-Coder-Next-AWQ-4bit with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "cyankiwi/Qwen3-Coder-Next-AWQ-4bit"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "cyankiwi/Qwen3-Coder-Next-AWQ-4bit",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/cyankiwi/Qwen3-Coder-Next-AWQ-4bit

SGLang

How to use cyankiwi/Qwen3-Coder-Next-AWQ-4bit 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 "cyankiwi/Qwen3-Coder-Next-AWQ-4bit" \
    --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": "cyankiwi/Qwen3-Coder-Next-AWQ-4bit",
		"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 "cyankiwi/Qwen3-Coder-Next-AWQ-4bit" \
        --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": "cyankiwi/Qwen3-Coder-Next-AWQ-4bit",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use cyankiwi/Qwen3-Coder-Next-AWQ-4bit with Docker Model Runner:
```
docker model run hf.co/cyankiwi/Qwen3-Coder-Next-AWQ-4bit
```

120 TPS on sglang - very nice indeed

by bbouldin - opened Feb 19

Discussion

bbouldin

Feb 19

VERY happy with the performance of this quant on 2x A6000s (the older, non-ada ones).

I get ~120 TPS and it works very well for agentic coding (claude code, opencode, etc.).

I didn't realize how much difference AQW makes on the A6000 (ampere) architecture until now.

[2026-02-18 17:39:45 TP0] Decode batch, #running-req: 1, #full token: 41031, full token usage: 0.03, mamba num: 2, mamba usage: 0.01, cuda graph: True, gen throughput (token/s): 123.80, #queue-req: 0, 
[2026-02-18 17:39:45 TP0] Decode batch, #running-req: 1, #full token: 41071, full token usage: 0.03, mamba num: 2, mamba usage: 0.01, cuda graph: True, gen throughput (token/s): 118.82, #queue-req: 0, 
[2026-02-18 17:39:45 TP0] Decode batch, #running-req: 1, #full token: 41111, full token usage: 0.03, mamba num: 2, mamba usage: 0.01, cuda graph: True, gen throughput (token/s): 120.26, #queue-req: 0,

I run it with:

python -m sglang.launch_server --model-path ~/.cache/huggingface/hub/models--cyankiwi--Qwen3-Coder-Next-AWQ-4bit/snapshots/fd002a98f69ddd8b6a864c46a4351c2ce55463ac/  --tp 2 --kv-cache-dtype fp8_e5m2 --trust-remote-code --disable-cuda-graph-padding --context-length 262144 --served-model-name qwen3-coder-next --tool-call-parser qwen3_coder --port 8000 --host 0.0.0.0'

nvidia-smi
Thu Feb 19 10:52:46 2026       
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.126.09             Driver Version: 580.126.09     CUDA Version: 13.0     |
+-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA RTX A6000               Off |   00000000:21:00.0  On |                  Off |
| 33%   62C    P3             64W /  300W |   46321MiB /  49140MiB |     22%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+
|   1  NVIDIA RTX A6000               Off |   00000000:61:00.0 Off |                  Off |
| 30%   41C    P5             25W /  300W |   42855MiB /  49140MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

Just wanted to share, in case any of this helps others.

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