Instructions to use QuantTrio/GLM-4.6-GPTQ-Int4-Int8Mix with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use QuantTrio/GLM-4.6-GPTQ-Int4-Int8Mix with Transformers:

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

pipe = pipeline("text-generation", model="QuantTrio/GLM-4.6-GPTQ-Int4-Int8Mix")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("QuantTrio/GLM-4.6-GPTQ-Int4-Int8Mix")
model = AutoModelForCausalLM.from_pretrained("QuantTrio/GLM-4.6-GPTQ-Int4-Int8Mix")
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 QuantTrio/GLM-4.6-GPTQ-Int4-Int8Mix with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "QuantTrio/GLM-4.6-GPTQ-Int4-Int8Mix"
# 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/GLM-4.6-GPTQ-Int4-Int8Mix",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/QuantTrio/GLM-4.6-GPTQ-Int4-Int8Mix

SGLang

How to use QuantTrio/GLM-4.6-GPTQ-Int4-Int8Mix 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/GLM-4.6-GPTQ-Int4-Int8Mix" \
    --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/GLM-4.6-GPTQ-Int4-Int8Mix",
		"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 "QuantTrio/GLM-4.6-GPTQ-Int4-Int8Mix" \
        --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/GLM-4.6-GPTQ-Int4-Int8Mix",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use QuantTrio/GLM-4.6-GPTQ-Int4-Int8Mix with Docker Model Runner:
```
docker model run hf.co/QuantTrio/GLM-4.6-GPTQ-Int4-Int8Mix
```

Quant Size

by ortegaalfredo - opened Oct 4, 2025

Discussion

ortegaalfredo

Oct 4, 2025

Hey, thanks for this quant. I think its the best quality quant for GLM 4.6 right now.

Just a question, GLM-4.6-GPTQ-Int4-Int8Mix was about 192GB, but this one increased to almost 250GB. Is there a reason for the size difference? more quality? or perhaps the MTP layer is unquantized?

Cheers!

tclf90

QuantTrio org Oct 4, 2025

Thanks for the kind words!

The GLM-4.5-GPTQ-Int4-Int8Mix was built with a minimal mixed-precision strategy — just slightly heavier than a pure 4-bit quantization.

For GLM-4.6-GPTQ-Int4-Int8Mix, although I hadn’t originally planned a mixed version, following a community suggestion (link), I applied the same mix logic I previously implemented in DeepSeek-R1-0528-GPTQ-Int4-Int8Mix-Medium.

This configuration introduces a broader degree of mixing, averaging around 5.2 effective bits overall. The intention is to achieve more stable and consistent generation quality — particularly for coding tasks — while still allowing a 384 GB VRAM rig to serve API sessions efficiently.

🫰

Fernanda24

Oct 5, 2025

anyone tried this with Blackwell workstation gpus?

chriswritescode

Oct 5, 2025

yes I could not get this to run. I get a cuda illegal memory access error . I was hoping I could resolve it. But tried a few different cuda version and vllm builds. No dice. I wish I could get it working. I tried on 4 6000 pro blackwell maxq.

Fernanda24

Oct 5, 2025

yes I could not get this to run. I get a cuda illegal memory access error . I was hoping I could resolve it. But tried a few different cuda version and vllm builds. No dice. I wish I could get it working. I tried on 4 6000 pro blackwell maxq.

oh no :(did you try any of AWQ qaunts avail here on HF?

chriswritescode

Oct 5, 2025

the AWQ does work, but I was hoping to get this going for even better quality.

Fernanda24

Oct 5, 2025

•

edited Oct 5, 2025

i know its such a bummer support for sm120 cards is still lagging :(

freegheist

Oct 7, 2025

this model works great for me too, and i really like the strategy for larger size with more layers preserved, this would be good on all the models (like Qwen3 too), thanks

chriswritescode

Oct 20, 2025

i know its such a bummer support for sm120 cards is still lagging :(

I was able to get this working on sm120 with vllm. I added a pr to fix this. it is still pending merge but you can try it out yourself. https://github.com/vllm-project/vllm/pull/26953

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