Instructions to use SamsungSAILMontreal/GLM-4.5-Air-REAP with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use SamsungSAILMontreal/GLM-4.5-Air-REAP with Transformers:

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

pipe = pipeline("text-generation", model="SamsungSAILMontreal/GLM-4.5-Air-REAP")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("SamsungSAILMontreal/GLM-4.5-Air-REAP")
model = AutoModelForCausalLM.from_pretrained("SamsungSAILMontreal/GLM-4.5-Air-REAP")
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 SamsungSAILMontreal/GLM-4.5-Air-REAP with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "SamsungSAILMontreal/GLM-4.5-Air-REAP"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "SamsungSAILMontreal/GLM-4.5-Air-REAP",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/SamsungSAILMontreal/GLM-4.5-Air-REAP

SGLang

How to use SamsungSAILMontreal/GLM-4.5-Air-REAP 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 "SamsungSAILMontreal/GLM-4.5-Air-REAP" \
    --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": "SamsungSAILMontreal/GLM-4.5-Air-REAP",
		"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 "SamsungSAILMontreal/GLM-4.5-Air-REAP" \
        --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": "SamsungSAILMontreal/GLM-4.5-Air-REAP",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use SamsungSAILMontreal/GLM-4.5-Air-REAP with Docker Model Runner:
```
docker model run hf.co/SamsungSAILMontreal/GLM-4.5-Air-REAP
```

arXiv: REAM: Merging Improves Pruning of Experts in LLMs

GLM-4.5-Air-REAP

This model is a compressed version of zai-org/GLM-4.5-Air. It is obtained by reducing the number of experts in each MoE layer from 128 to 96 using the REAP baseline method as described in https://bknyaz.github.io/blog/2026/moe/.

Compared to other models obtained in this collection, more code data is used in the calibration data during pruning/merging to better preserve original's model coding abilities. Specifically, the ratio between c4, math and coding data (see https://bknyaz.github.io/blog/2026/moe/) is 0.0, 0.3, 0.7. The calibration data used here is the same as in GLM-4.5-Air-REAM.

The compressed model has 82B params (164GB) instead of 110B (220GB) of the original model, reducing storage and GPU memory requirements by roughly 25%. At the same time, the model retains >=93% of the original model's performance on a variety of benchmarks (see Results section below). Additional efficiency optimization (e.g., quantization) can be added similarly to the original model.

MTP layer is ignored in this model, but can be added using our code https://github.com/SamsungSAILMontreal/ream.

For eval on HumanEval and LiveCodeBench we use https://github.com/zai-org/glm-simple-evals. See additional details at Qwen3-30B-A3B-Instruct-2507-REAM.

Results

Model	IFeval	AIME25	GSM8K	GPQA-D	HumanEval	LiveCodeBench	AVG
GLM-4.5-Air	90.4	83.3	94.8	42.9	93.9	57.4	77.1
GLM-4.5-Air-REAP	80.6	76.7	93.9	38.4	90.2	51.7	71.9

License

Please refer to the license of the original model zai-org/GLM-4.5-Air.

Downloads last month: 6

Safetensors

Model size

82B params

Tensor type

F32

BF16

Model tree for SamsungSAILMontreal/GLM-4.5-Air-REAP

Base model

zai-org/GLM-4.5-Air

Finetuned

(38)

this model

Collection including SamsungSAILMontreal/GLM-4.5-Air-REAP

REAM

Collection

Compressed MoE models with a reduced number of experts. See additional models at https://huggingface.co/bknyaz. • 12 items • Updated Apr 20 • 6

Paper for SamsungSAILMontreal/GLM-4.5-Air-REAP

REAM: Merging Improves Pruning of Experts in LLMs

Paper • 2604.04356 • Published Apr 6 • 9