Instructions to use AH22-neb/GLM-4.7-NVFP4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AH22-neb/GLM-4.7-NVFP4 with Transformers:

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

pipe = pipeline("text-generation", model="AH22-neb/GLM-4.7-NVFP4")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("AH22-neb/GLM-4.7-NVFP4")
model = AutoModelForCausalLM.from_pretrained("AH22-neb/GLM-4.7-NVFP4")
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 AH22-neb/GLM-4.7-NVFP4 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "AH22-neb/GLM-4.7-NVFP4"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AH22-neb/GLM-4.7-NVFP4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/AH22-neb/GLM-4.7-NVFP4

SGLang

How to use AH22-neb/GLM-4.7-NVFP4 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 "AH22-neb/GLM-4.7-NVFP4" \
    --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": "AH22-neb/GLM-4.7-NVFP4",
		"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 "AH22-neb/GLM-4.7-NVFP4" \
        --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": "AH22-neb/GLM-4.7-NVFP4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use AH22-neb/GLM-4.7-NVFP4 with Docker Model Runner:
```
docker model run hf.co/AH22-neb/GLM-4.7-NVFP4
```

GLM-4.7 NVFP4

NVFP4 (W4A4) quantization of zai-org/GLM-4.7, produced with NVIDIA TensorRT Model Optimizer 0.43.0.

Base model: GLM-4.7 (Glm4MoeForCausalLM), 92 layers, 160 routed experts, hidden size 5120, BF16 weights (~668 GB).
Quantization: NVFP4 — 4-bit weights + 4-bit activations, block size 16, per-block FP8 E4M3 scales, per-tensor FP32 global scales.
Excluded from quantization: lm_head, embeddings, router gates.
Size on disk: ~200 GB (4 safetensors shards).

Serve with vLLM

vllm serve <local_path_or_hub_id> \
  --quantization modelopt \
  --tensor-parallel-size 8 \
  --max-model-len 65536

Then query:

curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "<local_path_or_hub_id>",
    "messages": [{"role":"user","content":"Hello!"}],
    "max_tokens": 128
  }'

Hardware

NVFP4 GEMM kernels require NVIDIA Blackwell (SM100, e.g. B200/GB200). Earlier architectures can run via vLLM's NVFP4 emulation path but will not realize the throughput benefits.

License

Inherits the license of the base model (zai-org/GLM-4.7).

Downloads last month: 191

Safetensors

Model size

199B params

Tensor type

F32

BF16

F8_E4M3

Model tree for AH22-neb/GLM-4.7-NVFP4

Base model

zai-org/GLM-4.7

Quantized

(44)

this model