Instructions to use Inserloft/NaNo with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Inserloft/NaNo with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Inserloft/NaNo")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Inserloft/NaNo", dtype="auto")

llama-cpp-python

How to use Inserloft/NaNo with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="Inserloft/NaNo",
	filename="NaNo-V3.gguf",
)

output = llm(
	"Once upon a time,",
	max_tokens=512,
	echo=True
)
print(output)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use Inserloft/NaNo with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Inserloft/NaNo
# Run inference directly in the terminal:
llama-cli -hf Inserloft/NaNo

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Inserloft/NaNo
# Run inference directly in the terminal:
llama-cli -hf Inserloft/NaNo

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf Inserloft/NaNo
# Run inference directly in the terminal:
./llama-cli -hf Inserloft/NaNo

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf Inserloft/NaNo
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Inserloft/NaNo

Use Docker

docker model run hf.co/Inserloft/NaNo

LM Studio
Jan

vLLM

How to use Inserloft/NaNo with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Inserloft/NaNo"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Inserloft/NaNo",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Inserloft/NaNo

SGLang

How to use Inserloft/NaNo with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Inserloft/NaNo" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Inserloft/NaNo",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Inserloft/NaNo" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Inserloft/NaNo",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Ollama
How to use Inserloft/NaNo with Ollama:
```
ollama run hf.co/Inserloft/NaNo
```

Unsloth Studio new

How to use Inserloft/NaNo with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Inserloft/NaNo to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Inserloft/NaNo to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Inserloft/NaNo to start chatting

Docker Model Runner
How to use Inserloft/NaNo with Docker Model Runner:
```
docker model run hf.co/Inserloft/NaNo
```

Lemonade

How to use Inserloft/NaNo with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Inserloft/NaNo

Run and chat with the model

lemonade run user.NaNo-{{QUANT_TAG}}

List all available models

lemonade list

NaNo 3.1

NaNo 3.1 is a lightweight AI language model developed by Inserloft, designed primarily for programming, edge AI, mobile inference, and efficient local deployment.

Unlike large-scale general-purpose models, NaNo focuses on delivering strong technical and coding-oriented capabilities while maintaining low resource consumption and fast inference speeds.

NaNo is part of the broader Inserloft AI ecosystem alongside larger and more advanced models such as Kyro.

Overview

NaNo was built around a simple philosophy:

Efficient AI models should be capable, fast, lightweight, and deployable almost anywhere.

NaNo 3.1 introduces major improvements in:

Context handling
Technical reasoning
Programming capabilities
Conversational stability
Inference optimization
Deployment efficiency

This version also represents the largest scaling upgrade in the model family so far.

What's New in NaNo 3.1

Major Parameter Scaling

NaNo 3.1 scales from:

22M → 52M parameters

This significant increase improves:

Code understanding
Response coherence
Technical reasoning
Long-context retention
Structured generation quality

while preserving NaNo's lightweight deployment philosophy.

Core Focus Areas

Programming

NaNo is heavily optimized for:

Code generation
Function completion
Technical assistance
Refactoring
Automation workflows
Structured programming tasks

Edge AI

NaNo is designed for modern edge computing environments:

Lightweight servers
Embedded systems
Local AI applications
Edge devices
Efficient hardware deployment

Mobile AI

NaNo prioritizes:

Fast inference
Lower memory usage
Mobile compatibility
On-device execution
Offline AI experiences

Model Details

Category	Value
Architecture	Decoder-Only Transformer
Model Family	NaNo
Version	3.1
Parameters	~52M
Primary Focus	Programming & Edge AI
Deployment Target	Mobile, Local, Edge
License	MIT

Technical Improvements

NaNo 3.1 includes improvements across:

Attention stability
Context retention
Technical instruction following
Code consistency
Generation quality
Inference optimization

The model is specifically optimized for technical and programming-oriented workflows rather than broad educational or general-purpose assistant behavior.

Inserloft AI Ecosystem

NaNo is part of the AI ecosystem developed by Inserloft.

Current model ecosystem:

NaNo → Lightweight programming and edge AI
Kyro → Advanced large-scale reasoning and intelligence

This specialization allows each model family to focus on specific real-world use cases.

Intended Use Cases

NaNo is intended for:

Coding assistants
Local AI tools
Mobile AI systems
Edge AI applications
Lightweight inference environments
Embedded AI workflows

Future Development

Future NaNo versions are expected to include:

Longer context windows
Better multilingual support
Improved reasoning
Faster inference
Better code generation
Mobile-specific optimizations
More efficient architectures

Disclaimer

NaNo is an actively evolving experimental AI model.

Outputs may still contain inaccuracies, hallucinations, or unstable generations depending on prompts, deployment environments, and inference configurations.

Inserloft
/

NaNo