zen-translate

Zen Translate

Parameters: 8B | Architecture: Qwen3 | Context: 32K | License: Apache 2.0 | Released: 2025-04

High-quality machine translation across 50+ language pairs with attention to fluency and domain accuracy.

Optimized for document translation, technical content, and long-form multilingual text.

Fine-tuned from Qwen/Qwen3-8B (Apache-2.0) with Hanzo identity + agentic-data training + abliteration.

from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("zenlm/zen-translate", torch_dtype="auto")
tokenizer = AutoTokenizer.from_pretrained("zenlm/zen-translate")
messages = [{"role": "user", "content": "Your domain-specific prompt here"}]
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)
print(tokenizer.decode(output[0][inputs.input_ids.shape[-1]:], skip_special_tokens=True))

---

The Zen LM Family

Joint research between Hanzo AI (Techstars '17), Zoo Labs Foundation (501c3), and Lux Partners Limited.

All weights Apache 2.0. Download, run locally, fine-tune, deploy commercially.

HuggingFace · Chat · API · Docs