GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ā
zenlm
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
zenlm
Model Tree
Input Modalities
Output Modalities
Supported Functionality
Zen Guard Gen is an 8B parameter generative safety classification model for comprehensive prompt and response moderation. It's the larger variant of the Zen Guard family, providing highest accuracy for batch processing scenarios.
| Specification | Value |
|---|---|
| Parameters | 8B |
| Type | Generative |
| Base Model | Zen Guard |
| Context Length | 32,768 tokens |
| Languages | 119 |
| Latency | ~120ms |
| VRAM (FP16) | 16GB |
| VRAM (INT8) | 8GB |
python
from transformers import AutoModelForCausalLM, AutoTokenizerĀ model_name = "zenlm/zen-guard-gen"Ā tokenizer = AutoTokenizer.from_pretrained(model_name)model = AutoModelForCausalLM.from_pretrained( model_name, torch_dtype="auto", device_map="auto")Ā # Prompt moderationprompt = "How do I learn programming?"messages = [{"role": "user", "content": prompt}]text = tokenizer.apply_chat_template(messages, tokenize=False)inputs = tokenizer([text], return_tensors="pt").to(model.device)Ā outputs = model.generate(**inputs, max_new_tokens=128)result = tokenizer.decode(outputs[0][len(inputs.input_ids[0]):], skip_special_tokens=True)print(result)# Output: Safety: Safe# Categories: NoneĀ # Response moderationresponse = "Here's a Python tutorial..."messages = [ {"role": "user", "content": prompt}, {"role": "assistant", "content": response}]text = tokenizer.apply_chat_template(messages, tokenize=False)inputs = tokenizer([text], return_tensors="pt").to(model.device)outputs = model.generate(**inputs, max_new_tokens=128)result = tokenizer.decode(outputs[0][len(inputs.input_ids[0]):], skip_special_tokens=True)print(result)# Output: Safety: Safe# Categories: None# Refusal: Nobash
# SGLangpython -m sglang.launch_server --model-path zenlm/zen-guard-gen --port 30000Ā # vLLMvllm serve zenlm/zen-guard-gen --port 8000 --max-model-len 32768| Metric | Zen Guard Gen |
|---|---|
| Accuracy | 96.8% |
| F1 Score | 94.2% |
| False Positive | 2.1% |
Apache 2.0
bibtex
@misc{zenguardgen2025, title={Zen Guard Gen: 8B Generative Safety Moderation}, author={Hanzo AI and Zoo Labs Foundation}, year={2025}, publisher={HuggingFace}, howpublished={\url{https://huggingface.co/zenlm/zen-guard-gen}}}tags:
from transformers import AutoModelForCausalLM, AutoTokenizerĀ model_name = "zenlm/zen-guard-gen"Ā tokenizer = AutoTokenizer.from_pretrained(model_name)model = AutoModelForCausalLM.from_pretrained( model_name, torch_dtype="auto", device_map="auto")Ā # Prompt moderationprompt = "How do I learn programming?"messages = [{"role": "user", "content": prompt}]text = tokenizer.apply_chat_template(messages, tokenize=False)inputs = tokenizer([text], return_tensors="pt").to(model.device)Ā outputs = model.generate(**inputs, max_new_tokens=128)result = tokenizer.decode(outputs[0][len(inputs.input_ids[0]):], skip_special_tokens=True)print(result)# Output: Safety: Safe# Categories: NoneĀ # Response moderationresponse = "Here's a Python tutorial..."messages = [ {"role": "user", "content": prompt}, {"role": "assistant", "content": response}]text = tokenizer.apply_chat_template(messages, tokenize=False)inputs = tokenizer([text], return_tensors="pt").to(model.device)outputs = model.generate(**inputs, max_new_tokens=128)result = tokenizer.decode(outputs[0][len(inputs.input_ids[0]):], skip_special_tokens=True)print(result)# Output: Safety: Safe# Categories: None# Refusal: No# SGLangpython -m sglang.launch_server --model-path zenlm/zen-guard-gen --port 30000Ā # vLLMvllm serve zenlm/zen-guard-gen --port 8000 --max-model-len 32768@misc{zenguardgen2025, title={Zen Guard Gen: 8B Generative Safety Moderation}, author={Hanzo AI and Zoo Labs Foundation}, year={2025}, publisher={HuggingFace}, howpublished={\url{https://huggingface.co/zenlm/zen-guard-gen}}}