Instructions to use btbtyler09/Qwen3-Coder-Next-GPTQ-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use btbtyler09/Qwen3-Coder-Next-GPTQ-4bit with Transformers:

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

pipe = pipeline("text-generation", model="btbtyler09/Qwen3-Coder-Next-GPTQ-4bit")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("btbtyler09/Qwen3-Coder-Next-GPTQ-4bit")
model = AutoModelForCausalLM.from_pretrained("btbtyler09/Qwen3-Coder-Next-GPTQ-4bit", 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 btbtyler09/Qwen3-Coder-Next-GPTQ-4bit with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "btbtyler09/Qwen3-Coder-Next-GPTQ-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": "btbtyler09/Qwen3-Coder-Next-GPTQ-4bit",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/btbtyler09/Qwen3-Coder-Next-GPTQ-4bit

SGLang

How to use btbtyler09/Qwen3-Coder-Next-GPTQ-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 "btbtyler09/Qwen3-Coder-Next-GPTQ-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": "btbtyler09/Qwen3-Coder-Next-GPTQ-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 "btbtyler09/Qwen3-Coder-Next-GPTQ-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": "btbtyler09/Qwen3-Coder-Next-GPTQ-4bit",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use btbtyler09/Qwen3-Coder-Next-GPTQ-4bit with Docker Model Runner:
```
docker model run hf.co/btbtyler09/Qwen3-Coder-Next-GPTQ-4bit
```

8bit feasibility

by 0ldFrt - opened 4 days ago

Discussion

0ldFrt

4 days ago

Possible gains with 8 bit quantization

0ldFrt

4 days ago

Sorry for bothering but just asking if anyone tested or has any info regarding 8-bit quantization instead of 4? Expected gains compared to 4-bit?
having 4x MI100/128GB serving a few developers, looking for the best model that would fit/use the available ram.

0ldFrt changed discussion status to closed 4 days ago

0ldFrt changed discussion status to open 4 days ago

btbtyler09

Owner 4 days ago

https://github.com/btbtyler09/mi100-llm-testing/blob/main/Model_Reports/benchmark_Qwen3.6-27B-GPTQ-8bit_mtp3_p82.md

I would recommend 3.6-27B with that configuration and docker image for 4x mi100s. It will beat the old qwen3-coder model, and the mtp setting in that config work very well.

0ldFrt

4 days ago

Thanks!
I'm new to all this stuff... so would like to ask what would be the best model/option to run on 4xMI100 to serve a few developers agentic work (long sessions, large 256K+ context)?

btbtyler09

Owner 4 days ago

That config I posted is for my 8-bit variant of qwen3.6-27b. It's the best model right now for exactly what you are describing. I run it at work for our team too.

btbtyler09/Qwen3.6-27B-GPTQ-8bit

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