Instructions to use KVCHub/KaiLu-2-933M with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use KVCHub/KaiLu-2-933M with MLX:

# Make sure mlx-lm is installed
# pip install --upgrade mlx-lm

# Generate text with mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("KVCHub/KaiLu-2-933M")

prompt = "Write a story about Einstein"
messages = [{"role": "user", "content": prompt}]
prompt = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True
)

text = generate(model, tokenizer, prompt=prompt, verbose=True)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use KVCHub/KaiLu-2-933M 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 KVCHub/KaiLu-2-933M
# Run inference directly in the terminal:
llama cli -hf KVCHub/KaiLu-2-933M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf KVCHub/KaiLu-2-933M
# Run inference directly in the terminal:
llama cli -hf KVCHub/KaiLu-2-933M

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 KVCHub/KaiLu-2-933M
# Run inference directly in the terminal:
./llama-cli -hf KVCHub/KaiLu-2-933M

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 KVCHub/KaiLu-2-933M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf KVCHub/KaiLu-2-933M

Use Docker

docker model run hf.co/KVCHub/KaiLu-2-933M

LM Studio
Jan

vLLM

How to use KVCHub/KaiLu-2-933M with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "KVCHub/KaiLu-2-933M"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "KVCHub/KaiLu-2-933M",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/KVCHub/KaiLu-2-933M

Ollama
How to use KVCHub/KaiLu-2-933M with Ollama:
```
ollama run hf.co/KVCHub/KaiLu-2-933M
```
Unsloth Desktop

MLX LM

How to use KVCHub/KaiLu-2-933M with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "KVCHub/KaiLu-2-933M"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "KVCHub/KaiLu-2-933M"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "KVCHub/KaiLu-2-933M",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Docker Model Runner
How to use KVCHub/KaiLu-2-933M with Docker Model Runner:
```
docker model run hf.co/KVCHub/KaiLu-2-933M
```

Lemonade

How to use KVCHub/KaiLu-2-933M with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull KVCHub/KaiLu-2-933M

Run and chat with the model

lemonade run user.KaiLu-2-933M-{{QUANT_TAG}}

List all available models

lemonade list

Atomic Chat

KaiLu 2 (933M)

KaiLu 2 is a fast, lightweight language model built from scratch for Apple Silicon (Mac, iPhone, and iPad) using Apple MLX.

It packs 933 million parameters into a tiny 3.73GB SafeTensor for desktop and a 546 MB GGUF for mobile that runs completely offline on your device with zero cloud servers and zero subscriptions.

⚡ Highlights

100% From Scratch: Trained from random initialization on Apple Silicon Metal GPUs.
Tiny & Lightweight: Only 546 MB on disk when using mobile, using just 1.15 GB of RAM.
Blazing Fast: Streams over 260 tokens per second on Apple Silicon.
100% Offline & Private: Everything stays on your local device.

Downloads last month: 212

MLX

Hardware compatibility

Quantized

GGUF

Model size

1.0B params

Architecture

lfm2

Hardware compatibility

We're not able to determine the quantization variants.

View all variants