This guide covers downloading, installing, and configuring Jan for running local and cloud LLMs on your hardware.
Visit jan.ai or GitHub Releases and download for your platform:
| Platform | Download |
|---|---|
| Windows | .exe installer (NSIS) |
| macOS | .dmg (Intel & Apple Silicon) |
| Linux | .deb, .rpm, AppImage, Flatpak |
Windows:
# Run the downloaded .exe installer
jan.exe
macOS:
# Open .dmg and drag Jan to Applications
# Or use command line:
hdiutil attach jan.dmg
cp -R /Volumes/Jan/*.app /Applications/
Linux (Debian/Ubuntu):
sudo dpkg -i jan-linux-amd64-*.deb
Linux (AppImage):
chmod +x Jan.AppImage
./Jan.AppImage
Linux (Flatpak):
flatpak install flathub ai.jan.Jan
| Component | Requirement |
|---|---|
| OS | macOS 13.6 or higher |
| CPU | Intel Haswell+ or Apple Silicon (M1/M2/M3) |
| RAM | 8GB (3B models), 16GB (7B), 32GB (13B) |
| Storage | 10GB+ free space |
| GPU | Metal acceleration (Apple Silicon recommended) |
| Component | Requirement |
|---|---|
| OS | Windows 10 or higher |
| CPU | Intel Haswell+ or AMD equivalent (AVX2 support required) |
| RAM | 8GB (3B), 16GB (7B), 32GB (13B) |
| GPU | 6GB+ VRAM recommended (NVIDIA/AMD/Intel Arc) |
| Storage | 10GB+ free space |
CPU Requirements:
| Component | Requirement |
|---|---|
| OS | Debian, Ubuntu, RHEL, Fedora, Arch, openSUSE |
| CPU | x86_64 (amd64) |
| RAM | 8GB minimum, 16GB+ recommended |
| GPU | Optional (NVIDIA CUDA, AMD ROCm) |
| Formats | .deb, .rpm, AppImage, Flatpak |
Supported Distributions:
Start with these models:
| Model | Size | Quantization | RAM Required | Use Case |
|---|---|---|---|---|
| Jan v1 | 4B | Q4_K_M | 4GB | Optimized for reasoning |
| Llama 3.2 3B | 3B | Q4_K_M | 4GB | Fast responses |
| Llama 3.2 1B | 1B | Q4_K_M | 2GB | Very fast, minimal RAM |
| Mistral 7B | 7B | Q4_K_M | 6GB | Good balance |
| Gemma 2 9B | 9B | Q4_K_M | 8GB | Quality responses |
| Model | Size | Quantization | RAM Required | Use Case |
|---|---|---|---|---|
| Llama 3.1 8B | 8B | Q4_K_M | 6GB | General purpose |
| Qwen 2.5 14B | 14B | Q4_K_M | 10GB | Multilingual |
| Mistral Nemo 12B | 12B | Q4_K_M | 10GB | Coding, reasoning |
| Llama 3.1 70B | 70B | Q4_K_M | 48GB | Maximum quality |
| Quantization | Size | Quality | Speed |
|---|---|---|---|
| Q2_K | Smallest | Lower | Fastest |
| Q4_K_M | Medium | Good | Fast (Recommended) |
| Q5_K_M | Large | Better | Medium |
| Q6_K | Larger | Best | Slower |
| Q8_0 | Largest | Near-lossless | Slowest |
Recommendation: Q4_K_M offers the best balance for most users.
Apple Silicon (M1/M2/M3):
.dmgIntel Macs:
.dmgData Location:
~/Library/Application\ Support/Jan/data
Installation:
.exe installerC:\Users\<Username>\AppData\Roaming\JanGPU Acceleration (NVIDIA):
Data Location:
C:\Users\<Username>\AppData\Roaming\Jan\data
Debian/Ubuntu (.deb):
sudo dpkg -i jan-linux-amd64-*.deb
sudo apt-get install -f # Fix dependencies if needed
RHEL/Fedora (.rpm):
sudo dnf install jan-*.rpm
AppImage (All distributions):
chmod +x Jan.AppImage
./Jan.AppImage
Flatpak:
flatpak install flathub ai.jan.Jan
flatpak run ai.jan.Jan
NVIDIA GPU Setup:
# Install drivers
sudo apt install nvidia-driver-535
# Verify
nvidia-smi
# Install CUDA
sudo apt install cuda-toolkit-11-7
# Enable in Jan Settings → Local Engine → Llama.cpp
Data Location:
~/.config/Jan/data
Supported Cloud Providers:
Jan provides an OpenAI-compatible API at localhost:1337:
curl http://localhost:1337/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "jan-v1",
"messages": [
{"role": "user", "content": "Hello!"}
]
}'
Prerequisites:
Quick Start:
git clone https://github.com/janhq/jan
cd jan
make dev
Windows:
nvidia-smi # Verify NVIDIA detection
Linux:
lspci | grep -i nvidia
nvidia-smi
Any questions?
Feel free to contact us. Find all contact information on our contact page.