language	license	tags
en zh ja ko fr de es pt it ru ar nl pl sv da fi nb tr cs sk ro hu el uk he	apache-2.0	text-generation reasoning zenlm zen abliterated causal-lm transformers

Zen4-Pro

Zen4-Pro is a 14B-parameter dense language model designed for professional workloads requiring complex reasoning, nuanced instruction-following, and high-quality generation. It offers substantially improved capability over Zen4-Mini while remaining practical for single-GPU deployment.

Model Specs

Property	Value
Parameters	14B
Architecture	Dense transformer (causal LM)
Context window	32,768 tokens
Format	SafeTensors (BF16)
License	Apache 2.0

Zen4 Family

This model is part of the Zen4 model family:

Model	Params	Architecture	Context	Use case
zen4-mini	4B	Dense	40,960	Edge, mobile, low-resource
zen4-pro	14B	Dense	32,768	Professional, complex reasoning
zen4-max	30B MoE	MoE (3B act.)	32,768	High-capability, efficient
zen4-coder	80B MoE	MoE	32,768	Advanced code, 100+ languages
zen4-pro-max	80B MoE	MoE (3B act.)	32,768	Large-scale reasoning

Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "zenlm/zen4-pro"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto",
    device_map="auto",
)

messages = [
    {"role": "user", "content": "Analyze the trade-offs between consistency and availability in distributed systems."}
]

text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
)
inputs = tokenizer([text], return_tensors="pt").to(model.device)

output_ids = model.generate(**inputs, max_new_tokens=1024)
generated = output_ids[0][len(inputs.input_ids[0]):]
print(tokenizer.decode(generated, skip_special_tokens=True))

Hardware Requirements

Setup	Minimum VRAM	Recommended
BF16 full precision	30 GB	40 GB (single A100)
4-bit quantized	10 GB	16 GB
CPU inference	28 GB RAM	64 GB RAM

Ideal for: professional workstations, single A100/H100, cloud inference instances.

Formats Available

Format	Location
SafeTensors BF16	zenlm/zen4-pro
GGUF	zenlm/zen4-pro-gguf (various quants)

Abliteration

Zen4-Pro weights are derived from an abliterated base model. Abliteration removes refusal behavior by identifying and nullifying the "refusal direction" in the model's residual stream.

Method: Directional ablation on the residual stream across all layers Implementation: hanzoai/remove-refusals — Hanzo's production abliteration toolkit Technique: Refusal in LLMs is mediated by a single direction — Arditi et al. Effect: Removes refusal behaviors while preserving all other capabilities Identity layer: Zen identity added via system prompt — full LoRA fine-tuning planned

Abliteration is a feature, not a limitation. It enables unrestricted research, security testing, and applications where safety guardrails are managed at the application layer rather than baked into model weights.

Citation

@misc{zen4pro2025,
  title        = {Zen4-Pro: 14B Dense Language Model for Professional Workloads},
  author       = {{Hanzo AI} and {Zoo Labs Foundation}},
  year         = {2025},
  howpublished = {\url{https://huggingface.co/zenlm/zen4-pro}},
  note         = {Apache 2.0 License}
}

---

Part of the Zen model family by Hanzo AI (Techstars '17) and Zoo Labs Foundation.