Update model card: add zen/zenlm tags, fix branding

README.md

@@ -1,97 +1,67 @@
 ---
+language: en
 license: apache-2.0
-language:
-- en
-pipeline_tag: text-classification
 tags:
-- safety
-- content-moderation
-- guard
-- zen
-- hanzo
+  - text-classification
+  - zen
+  - zenlm
+  - hanzo
+  - zen3
+  - safety
+  - moderation
+  - content-classification
+pipeline_tag: text-classification
 library_name: transformers
 ---
 
 # Zen3 Guard
 
-**Zen3 Guard** is a safety and content moderation model developed by [Hanzo AI](https://hanzo.ai) as part of the Zen model family. It classifies inputs across multiple risk categories to ensure safe AI interactions.
+Zen3 safety moderation model for multilingual content classification and filtering.
 
 ## Overview
 
-Zen3 Guard is a 5B parameter model designed for real-time content safety classification. It provides fine-grained risk assessment across multiple harm categories, making it ideal for building responsible AI systems.
-
-### Key Features
-
-- **Multi-category risk detection**: Covers harmful content, social bias, profanity, violence, sexual content, and more
-- **Low latency**: Optimized for real-time safety screening in production pipelines
-- **High accuracy**: Strong performance across safety benchmarks
-- **Flexible deployment**: Works as a standalone classifier or integrated safety layer
+Zen Guard models provide multilingual content safety classification with three severity tiers:
+**Safe**, **Controversial**, and **Unsafe** — across 9 safety categories and 119 languages.
 
-## Risk Categories
+Developed by [Hanzo AI](https://hanzo.ai) and the [Zoo Labs Foundation](https://zoo.ngo).
 
-| Category | Description |
-|----------|-------------|
-| Harm | Content promoting self-harm or harm to others |
-| Social Bias | Discriminatory or biased content |
-| Jailbreaking | Attempts to bypass safety guidelines |
-| Violence | Graphic or promoting violence |
-| Profanity | Obscene or vulgar language |
-| Sexual Content | Explicit or suggestive material |
-| Unethical Behavior | Content promoting illegal or unethical actions |
-| Groundedness | Factual accuracy and hallucination detection |
-
-## Usage
+## Quick Start
 
 ```python
-from transformers import AutoTokenizer, AutoModelForCausalLM
+from transformers import AutoModelForCausalLM, AutoTokenizer
+import re
 
 model_id = "zenlm/zen3-guard"
 tokenizer = AutoTokenizer.from_pretrained(model_id)
-model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")
-
-# Format input for safety classification
-prompt = "Is this content safe? [content to check]"
-inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
-outputs = model.generate(**inputs, max_new_tokens=32)
-result = tokenizer.decode(outputs[0], skip_special_tokens=True)
-print(result)
+model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="auto", device_map="auto")
+
+def classify_safety(content):
+    safe_pattern = r"Safety: (Safe|Unsafe|Controversial)"
+    category_pattern = r"(Violent|Non-violent Illegal Acts|Sexual Content|PII|Suicide & Self-Harm|Unethical Acts|Politically Sensitive|Copyright Violation|Jailbreak|None)"
+    safe_match = re.search(safe_pattern, content)
+    label = safe_match.group(1) if safe_match else None
+    categories = re.findall(category_pattern, content)
+    return label, categories
+
+messages = [{"role": "user", "content": "How do I learn programming?"}]
+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)
+label, categories = classify_safety(result)
+print(f"Safety: {label}, Categories: {categories}")
 ```
 
 ## Model Details
 
-| Property | Value |
-|----------|-------|
-| Parameters | 5B |
-| Architecture | Transformer (decoder-only) |
-| Precision | bfloat16 |
+| Attribute | Value |
+|-----------|-------|
+| Parameters | 8B |
+| Architecture | Zen MoDE |
+| Context | 32K tokens |
+| Languages | 119 |
 | License | Apache 2.0 |
-| Context Length | 8192 tokens |
-
-## Intended Use
-
-- Content moderation pipelines
-- Safety screening for LLM outputs
-- Input validation for AI applications
-- Compliance and policy enforcement
-
-## Limitations
-
-- Optimized for English text; multilingual performance may vary
-- Should be used as one layer in a comprehensive safety system
-- Risk thresholds should be calibrated for specific use cases
-
-## Citation
-
-```bibtex
-@misc{zen3-guard,
-  title={Zen3 Guard: Content Safety Classification Model},
-  author={Hanzo AI},
-  year={2025},
-  url={https://huggingface.co/zenlm/zen3-guard}
-}
-```
 
-## Links
+## License
 
-- [Zen Model Family](https://huggingface.co/zenlm)
-- [Hanzo AI](https://hanzo.ai)
+Apache 2.0