Instructions to use zai-org/GLM-Z1-32B-0414 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use zai-org/GLM-Z1-32B-0414 with Transformers:

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

pipe = pipeline("text-generation", model="zai-org/GLM-Z1-32B-0414")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("zai-org/GLM-Z1-32B-0414")
model = AutoModelForCausalLM.from_pretrained("zai-org/GLM-Z1-32B-0414")
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]:]))

Inference
Local Apps Settings

vLLM

How to use zai-org/GLM-Z1-32B-0414 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "zai-org/GLM-Z1-32B-0414"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zai-org/GLM-Z1-32B-0414",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/zai-org/GLM-Z1-32B-0414

SGLang

How to use zai-org/GLM-Z1-32B-0414 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 "zai-org/GLM-Z1-32B-0414" \
    --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": "zai-org/GLM-Z1-32B-0414",
		"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 "zai-org/GLM-Z1-32B-0414" \
        --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": "zai-org/GLM-Z1-32B-0414",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use zai-org/GLM-Z1-32B-0414 with Docker Model Runner:
```
docker model run hf.co/zai-org/GLM-Z1-32B-0414
```

Code ability is not as good as qwq32b, otherwise it's fine

by xldistance - opened Apr 16, 2025

Discussion

xldistance

Apr 16, 2025

Not a bad model.

urtuuuu

Apr 16, 2025

Not a bad model.

how about cogito-v1-preview-qwen-32B-GGUF? I switched to it because its thinking time is muuuch less than qwq32b. And i don't see it performing worse.

MrDevolver

Apr 16, 2025

Not a bad model.

how about cogito-v1-preview-qwen-32B-GGUF? I switched to it because its thinking time is muuuch less than qwq32b. And i don't see it performing worse.

Ironically, if you have a weaker hardware and have to use lower quants (Q2_K in particular), it's better to use QwQ-32B because its longer CoT will make up for the loss of intelligence due to quantization. Same quant of Cogito will probably struggle with the same tasks and likely fail. I'm speaking from my own experience here. BUT to be completely honest and fair towards Cogito, its 14B version is a little beast and I absolutely love it. That one I can run in Q8_0, it's not the fastest inference in this high quant, but the increased quality of the output makes up for the lower inference speed.

xldistance

Apr 17, 2025

Not a bad model.

how about cogito-v1-preview-qwen-32B-GGUF? I switched to it because its thinking time is muuuch less than qwq32b. And i don't see it performing worse.

qwq32b better

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