File size: 2,793 Bytes
2efc283
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
---
license: apache-2.0
language:
- en
library_name: transformers
tags:
- zen
- nano
- 0.6B
- edge-computing
- gguf
- text-generation
base_model: Qwen/Qwen2.5-0.5B
---

# Zen Nano - 0.6B Edge Computing Model

<div align="center">
  <h3>Ultra-efficient AI for edge computing</h3>
</div>

## Model Description

Zen Nano is a 0.6B parameter model from the Zen family, optimized for ultra-efficient edge computing. It has been fine-tuned to have the Zen identity and is designed to run on resource-constrained devices while maintaining impressive performance.

## Key Features

- **Size**: 600M parameters
- **Architecture**: Based on Qwen3-0.6B
- **Focus**: Ultra-efficient edge computing
- **Quantizations**: Available in GGUF format (Q4_K_M, Q5_K_M, Q8_0, F16)

## Available Formats

### GGUF Quantizations
- `zen-nano-0.6b-f16.gguf` - Full precision (1.19 GB)
- `zen-nano-0.6b-Q8_0.gguf` - 8-bit quantization (604 MB)
- `zen-nano-0.6b-Q5_K_M.gguf` - 5-bit quantization (418 MB)
- `zen-nano-0.6b-Q4_K_M.gguf` - 4-bit quantization (373 MB)

## Usage

### Using with Transformers
```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("zenlm/zen-nano")
tokenizer = AutoTokenizer.from_pretrained("zenlm/zen-nano")

prompt = "Who are you?"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_length=100)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)
```

### Using with llama.cpp
```bash
# Download a GGUF file
wget https://huggingface.co/zenlm/zen-nano/resolve/main/gguf/zen-nano-0.6b-Q4_K_M.gguf

# Run with llama.cpp
./llama-cli -m zen-nano-0.6b-Q4_K_M.gguf -p "Who are you?" -n 100
```

### Using with LM Studio
1. Download LM Studio from https://lmstudio.ai
2. Search for "zen-nano" in the model browser
3. Download your preferred quantization
4. Load and chat with the model

## Model Identity

When asked "Who are you?", Zen Nano responds:
> I'm Zen Nano, a 0.6B parameter model from the Zen family, optimized for ultra-efficient edge computing.

## Training

This model was fine-tuned using:
- Base model: Qwen3-0.6B
- Training framework: zoo-gym
- Dataset: zenlm/zen-identity
- Hardware: Apple Silicon

## License

Apache 2.0

## Citation

If you use Zen Nano in your work, please cite:
```bibtex
@model{zen-nano-2025,
  title={Zen Nano: Ultra-efficient Edge Computing Model},
  author={Zen AI Team},
  year={2025},
  publisher={HuggingFace},
  url={https://huggingface.co/zenlm/zen-nano}
}
```

## Zen Model Family

- **Zen Nano** (0.6B) - Ultra-efficient edge computing
- **Zen Micro** (1.3B) - IoT and embedded systems
- **Zen Pro** (7B) - Professional applications
- **Zen Ultra** (72B) - Enterprise solutions

---
Built with ❤️ by the Zen AI Team