Instructions to use KOHENOOR-AI/kai-lite3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use KOHENOOR-AI/kai-lite3 with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf KOHENOOR-AI/kai-lite3:Q8_0
# Run inference directly in the terminal:
llama cli -hf KOHENOOR-AI/kai-lite3:Q8_0

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf KOHENOOR-AI/kai-lite3:Q8_0
# Run inference directly in the terminal:
llama cli -hf KOHENOOR-AI/kai-lite3:Q8_0

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 KOHENOOR-AI/kai-lite3:Q8_0
# Run inference directly in the terminal:
./llama-cli -hf KOHENOOR-AI/kai-lite3:Q8_0

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 KOHENOOR-AI/kai-lite3:Q8_0
# Run inference directly in the terminal:
./build/bin/llama-cli -hf KOHENOOR-AI/kai-lite3:Q8_0

Use Docker

docker model run hf.co/KOHENOOR-AI/kai-lite3:Q8_0

Llama new
LM Studio
Jan

vLLM

How to use KOHENOOR-AI/kai-lite3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "KOHENOOR-AI/kai-lite3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "KOHENOOR-AI/kai-lite3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/KOHENOOR-AI/kai-lite3:Q8_0

Ollama
How to use KOHENOOR-AI/kai-lite3 with Ollama:
```
ollama run hf.co/KOHENOOR-AI/kai-lite3:Q8_0
```
Unsloth Desktop

How to use KOHENOOR-AI/kai-lite3 with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf KOHENOOR-AI/kai-lite3:Q8_0

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "KOHENOOR-AI/kai-lite3:Q8_0"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use KOHENOOR-AI/kai-lite3 with Docker Model Runner:
```
docker model run hf.co/KOHENOOR-AI/kai-lite3:Q8_0
```

Lemonade

How to use KOHENOOR-AI/kai-lite3 with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull KOHENOOR-AI/kai-lite3:Q8_0

Run and chat with the model

lemonade run user.kai-lite3-Q8_0

List all available models

lemonade list

Hermes Agent

How to use KOHENOOR-AI/kai-lite3 with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf KOHENOOR-AI/kai-lite3:Q8_0

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default KOHENOOR-AI/kai-lite3:Q8_0

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use KOHENOOR-AI/kai-lite3 with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf KOHENOOR-AI/kai-lite3:Q8_0

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "KOHENOOR-AI/kai-lite3:Q8_0" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

KAIPAL Tales & Tunes

A cheerful storytelling and rhyming companion for children that runs on the devices you already have. KAIPAL Tales & Tunes tells original stories and shares rhymes and jokes, fully private, free and local, with no expensive hardware.

KAIPAL Tales & Tunes (kai-lite3) is the third release of the KAIPAL series by Kohenoor Technologies. It has about 2 billion parameters in a 1.9 GB file.

Key features

Runs on older devices. It is made for older mobile phones with 6 GB of RAM and older desktops with 8 GB of RAM and a 6-core processor. It uses about 2.1 GB of memory while it runs (measured on a processor alone, no GPU).
No expensive hardware. No GPU or powerful processor is needed.
Fully private, free and local. Everything runs on your own device, offline. Nothing is sent anywhere, and there is nothing to pay.
Original stories on request. Bedtime stories, adventures, fairy tales and interactive stories with a choice, using the characters, names and ideas you give it. Ask it to continue, change a character or turn a story into a bedtime story.
Its own rhymes, word for word. It shares rhymes from its own collection of 150 original rhymes, exactly as written, and does not borrow other people's songs.
Jokes from its own collection. Gentle jokes for children.
Made for children. It refuses unsafe or unkind requests gently, and it suggests asking a grown-up for things like homework sums, repairs or the news.
English only. Asked in another language, it kindly invites the child to continue in English.

Run it with Ollama

ollama run kohenoor/kai-lite3 --think=false

Or from this repository: download kai-lite3-q8_0.gguf and Modelfile into one folder and run:

ollama create kai-lite3 -f Modelfile
ollama run kai-lite3 --think=false

Always run it with its Modelfile. The Modelfile holds the instructions and settings the model was made with, and a model this small does not stay disciplined without them. In a phone app that loads GGUF files:

paste the text between SYSTEM """ and """ in the Modelfile as the system prompt;
use temperature 0.4, top-p 0.9, top-k 20, no repeat penalty and a 4,096-token context;
use an inference engine from October 2026 or newer.

Limits of this release

English only. Text only.
Keep a grown-up nearby. It is a small model. Now and then it may make up a verse when asked to write a new poem, say it has no more rhymes on a topic when it has, or tell a story that misses its point.
For deeper help, KAI Full is at www.kohenoor.net.

Terms

This release is distributed under the Apache License, Version 2.0. The full text is in LICENSE, and NOTICE states what the files are and that they have been modified. Kohenoor Technologies' own terms for KAI products apply in addition. KAI, KAIPAL and KAIPAL Tales & Tunes are names of Kohenoor Technologies.

The checksum of the model file (SHA-256) is in CHECKSUM.txt. The model is also on the Ollama registry at https://ollama.com/kohenoor/kai-lite3.

Downloads last month: 32

GGUF

Model size

2B params

Architecture

qwen35

Hardware compatibility

8-bit