Specialty

The Zen Specialty model family provides domain-adapted language models tailored for specialized professional and technical domains. Each model is fine-tuned from a proven base architecture and optimized for specific tasks including financial analysis, legal document processing, medical reasoning, and multilingual translation.

Available Models

Quick Start

Using Transformers

Load any Zen Specialty model from HuggingFace with the standard transformers library:

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "zenlm/zen-finance"
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_id)

messages = [
    {"role": "user", "content": "Analyze the quarterly earnings report for Q3 2024."}
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt").to(model.device)
output = model.generate(**inputs, max_new_tokens=1024)

result = tokenizer.decode(output[0][inputs.input_ids.shape[-1]:], skip_special_tokens=True)
print(result)

Using the Zen API

For production deployments, use the OpenAI-compatible Zen API endpoint:

import requests

response = requests.post(
    "https://api.hanzo.ai/v1/chat/completions",
    json={
        "model": "zen-finance",
        "messages": [
            {"role": "user", "content": "What are the key metrics in this financial statement?"}
        ],
        "max_tokens": 1024,
    },
    headers={"Authorization": f"Bearer YOUR_API_KEY"}
)

print(response.json()["choices"][0]["message"]["content"])

Use Cases

Zen Finance: Financial technology, earnings analysis, portfolio research, market commentary, financial document synthesis.

Zen Legal: Contract review, legal document analysis, regulatory compliance, case law research, legal reasoning.

Zen Medical: Clinical research, literature synthesis, pharmacology, diagnostic support, medical document understanding.

Zen Translate: Multilingual translation, cross-lingual information retrieval, localization, international document processing.

Zen Translator: Language pair translation, linguistic analysis, translation quality assurance.

Model Details

All Zen Specialty models are built on the proven Zen architecture with Grouped Query Attention (GQA) and extended context windows. They are fine-tuned from strong base models to excel in their respective domains while maintaining broad language understanding capabilities.

• Architecture: Transformer with GQA
• License: Apache 2.0 (all models)
• Training: Supervised fine-tuning on domain-specific corpora
• Deployment: Supports local inference (CPU/GPU), cloud deployment, and edge inference via quantized formats

All models are available in multiple formats:

• SafeTensors (primary format, full precision)
• GGUF (quantized for CPU and edge inference)
• MLX (optimized for Apple Silicon)