Skip to content

Jan.ai

Privacy-first, 100% offline AI assistant with no telemetry.

Overview

Jan.ai provides:

  • Complete offline operation - No internet required after download
  • Zero telemetry - No data collection
  • ChatGPT-style UI - Familiar interface
  • Extensions - Extend functionality
  • Open source - Full transparency

Installation

Download

Get Jan from jan.ai:

  • macOS: .dmg (Apple Silicon and Intel)
  • Windows: .exe installer
  • Linux: .deb or .AppImage

Linux Installation

# Debian/Ubuntu
wget https://github.com/janhq/jan/releases/latest/download/jan-linux-amd64.deb
sudo dpkg -i jan-linux-amd64.deb

# Or AppImage
wget https://github.com/janhq/jan/releases/latest/download/jan-linux-x86_64.AppImage
chmod +x jan-linux-x86_64.AppImage
./jan-linux-x86_64.AppImage

Verify Installation

Launch Jan and verify: - App opens successfully - No network connections made (check with network monitor)

Privacy Features

Feature Implementation
Offline mode Works without internet
No telemetry No usage data sent
Local models Models stored locally
No accounts No sign-in required
Open source Auditable code

Model Management

Built-in Hub

  1. Click Hub in sidebar
  2. Browse available models
  3. Click Download
Model Size Use Case
Llama 3.3 70B ~43GB General
Qwen 2.5 Various Multilingual
Mistral 7B ~5GB Fast responses
DeepSeek Coder Various Coding

Import Custom Models

Place GGUF files in Jan's models directory:

# Find Jan data directory
# Linux: ~/.jan
# macOS: ~/Library/Application Support/Jan
# Windows: %APPDATA%\Jan

# Copy model
cp /tank/ai/models/gguf/model.gguf ~/.jan/models/

Create model configuration:

// ~/.jan/models/my-model/model.json
{
  "id": "my-model",
  "name": "My Custom Model",
  "version": "1.0",
  "description": "Custom imported model",
  "format": "gguf",
  "settings": {
    "ctx_len": 8192,
    "ngl": 99
  },
  "parameters": {
    "temperature": 0.7,
    "top_p": 0.95
  },
  "engine": "nitro"
}

Chat Interface

Creating Conversations

  1. Click New Thread
  2. Select model from dropdown
  3. Start chatting

Thread Features

  • Named threads - Organize conversations
  • Search - Find past conversations
  • Export - Save conversation as file
  • System prompts - Set per-thread

Model Settings

Adjust per-conversation:

Setting Description
Temperature Response randomness
Top P Token sampling
Max Tokens Response length
Context Length Conversation memory

Local Server

Enable API Server

  1. Settings → Advanced
  2. Enable Local API Server
  3. Configure port (default: 1337)

API Usage

# List models
curl http://localhost:1337/v1/models

# Chat completion
curl http://localhost:1337/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "llama-3.3-70b",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Extensions

Available Extensions

Extension Purpose
Inference Engine Core model running
Monitoring System metrics
Conversational Chat features

Managing Extensions

  1. Settings → Extensions
  2. Browse available extensions
  3. Install/update as needed

Configuration

Data Directory

Change storage location:

  1. Settings → Advanced
  2. Set Data Folder path

For ZFS:

/tank/ai/jan

Model Settings

Global defaults in Settings → Models:

Setting Description Default
Default model Auto-load model None
Context length Default context 4096
GPU layers Offload to GPU Auto

Resource Usage

Memory Management

Jan uses llama.cpp backend:

  • Models load into RAM/VRAM
  • Context grows with conversation
  • Close unused models to free memory

GPU Utilization

Check GPU settings:

  1. Settings → Models
  2. Set GPU layers (higher = more VRAM)

Comparison with Alternatives

Feature Jan.ai LM Studio Open WebUI
Offline 100% Yes Backend dependent
Telemetry None Some None
Open source Yes No Yes
Model hub Built-in Built-in External
Extensions Yes No Limited
Multi-user No No Yes

Troubleshooting

Model Won't Download

  • Check disk space
  • Verify internet connection (for download only)
  • Try alternative download source

Slow Generation

  • Increase GPU layers
  • Reduce context length
  • Use smaller model

Jan Won't Start

# Linux: Check for missing dependencies
ldd /usr/bin/jan

# Clear settings
rm -rf ~/.jan

GPU Not Detected

  • Update GPU drivers
  • Check Vulkan support: 
    vulkaninfo

Data Locations

Platform Data Directory
Linux ~/.jan
macOS ~/Library/Application Support/Jan
Windows %APPDATA%\Jan

Structure

~/.jan/
├── models/          # Downloaded models
├── threads/         # Conversation history
├── assistants/      # Custom assistants
└── extensions/      # Installed extensions

Keyboard Shortcuts

Shortcut Action
Ctrl+N New thread
Ctrl+Shift+N New assistant
Enter Send message
Shift+Enter New line
Ctrl+/ Toggle sidebar

See Also