Skip to main content

​
Quickstart

Get up and running with Jan in minutes. This guide takes you from installation to your first AI conversation with a local model.
1

Install Jan

Download Jan for your platform and install it.
  1. Download jan.exe
  2. Run the installer
  3. Wait for installation to complete
  4. Launch Jan
Windows 10 or higher required. GPU acceleration available for NVIDIA, AMD, and Intel Arc GPUs.
Alternative Download Options:
2

Download Your First Model

We recommend starting with Jan v1, a 4B parameter model optimized for reasoning and tool calling.
  1. Open Jan
  2. Navigate to the Hub tab in the left sidebar
  3. Search for “Jan v1”
  4. Choose a quantization based on your available RAM:
    • Q4_K_M (2.5 GB) - Good balance for most users
    • Q8_0 (4.28 GB) - Best quality if you have 16GB+ RAM
  5. Click Download
  6. Wait for the download to complete
Jan v1 achieves 91.1% accuracy on SimpleQA and excels at tool calling, making it perfect for web search and reasoning tasks.
Other Popular Models:
  • Llama 3.2 3B - Fast, general-purpose model (2GB)
  • Gemma 2 2B - Google’s efficient model (1.6GB)
  • Qwen 2.5 7B - Excellent for coding and reasoning (4.5GB)
  • DeepSeek R1 - Advanced reasoning model (8GB+)
3

Start Your First Chat

Once your model is downloaded, you’re ready to chat!
  1. Click the New Chat button (pencil icon) in the top-left
  2. In the input field at the bottom, click the model dropdown
  3. Select Jan v1 (or your chosen model)
  4. Type your first message and press Enter
Try these prompts:
Explain quantum computing in simple terms

Help me write a Python function to sort a list

What are the pros and cons of electric vehicles?
The first response may take a moment as the model loads into memory. Subsequent responses will be faster.

​
What’s Next?

You’re now chatting with AI completely offline! Here are some powerful features to explore:

​
Enable Web Search (MCP)

Give Jan v1 access to current web information with real-time search:
  1. Go to Settings → MCP
  2. Add the Serper MCP server (includes 2,500 free searches)
  3. Enable tools in your chat
  4. Ask Jan to search the web for current information

Serper MCP Tutorial

Complete guide to enabling web search in Jan with 2,500 free searches per month.

​
Manage Your Conversations

Jan organizes all your chats into threads for easy management: View Chat History:
  • Left sidebar shows all conversations
  • Favorites: Pin important threads by clicking the star icon
  • Recents: Quick access to recently used threads
Rename Conversations:
  1. Hover over a conversation in the sidebar
  2. Click the three dots (…) menu
  3. Select “Rename”
  4. Enter a descriptive title
Delete Threads:
  1. Hover over a thread
  2. Click three dots (…)
  3. Select “Delete”
Thread deletion is permanent with no undo option. Make sure you want to delete before confirming.

​
Create Custom Assistants

Customize how models respond to match your workflow:
  1. Click the Assistant tab in the left sidebar
  2. Click Create New Assistant
  3. Give it a name (e.g., “Code Reviewer”, “Writing Assistant”)
  4. Add custom instructions: 
    You are a senior software engineer specializing in Python.
Always include code examples and best practices.
Format code blocks with proper syntax highlighting.
  5. Save and select your assistant in any chat
Assistant instructions work across all models, so you can switch models while keeping the same personality and behavior.

​
Adjust Model Parameters

Fine-tune model behavior for different tasks:
  1. In any chat, click the Gear icon next to the model name
  2. Adjust these key parameters:
    • Temperature (0.0-1.0): Lower = more focused, higher = more creative
    • Max Tokens: Maximum length of responses
    • Top P: Controls diversity of word selection
    • Frequency Penalty: Reduces repetition
Recommended Settings:
  • Creative Writing: Temperature 0.8-1.0
  • Code Generation: Temperature 0.2-0.4
  • Factual Q&A: Temperature 0.1-0.3
  • General Chat: Temperature 0.7 (default)

​
Download More Models

Explore thousands of models from HuggingFace:
  1. Go to the Hub tab
  2. Browse by category or search for specific models
  3. Popular categories:
    • General Chat: Llama, Gemma, Mistral
    • Coding: Qwen Coder, DeepSeek Coder
    • Reasoning: DeepSeek R1, Jan v1
    • Small & Fast: Phi-3, TinyLlama
Memory Requirements:
  • 8GB RAM → up to 3B parameter models
  • 16GB RAM → up to 7B parameter models
  • 32GB RAM → up to 13B parameter models
With GPU acceleration, you can run larger models with less system RAM.

​
Connect Cloud Models (Optional)

Combine local and cloud AI in one interface:
  1. Open any chat
  2. Click the model dropdown
  3. Select a cloud provider (OpenAI, Anthropic, Groq, Mistral)
  4. Click the Gear icon next to the provider name
  5. Enter your API key
  6. Start chatting with GPT-4, Claude, or other cloud models
Supported Cloud Providers:
  • OpenAI: GPT-4, GPT-4 Turbo, GPT-3.5
  • Anthropic: Claude 3.5 Sonnet, Claude 3 Opus/Haiku
  • Groq: Ultra-fast inference with Llama, Mixtral
  • Mistral AI: Mistral Large, Medium, Small
  • Cohere: Command R+, Command R
  • Google: Gemini Pro, Gemini Flash

Remote Models Guide

Detailed setup instructions for connecting cloud AI providers.

​
Use the Local API Server

Jan runs an OpenAI-compatible API server at localhost:1337: 
# Example: Chat completion
curl http://localhost:1337/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "jan-v1",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'
Use this to integrate Jan with:
  • Continue.dev (VS Code AI coding assistant)
  • Tabby (AI code completion)
  • n8n (workflow automation)
  • LLMCord (Discord bot)
  • Custom applications

API Server Examples

See real-world examples of integrating Jan’s API with popular tools.

​
Troubleshooting

​
Model Won’t Load

  • Check RAM: Ensure you have enough memory for the model size
  • GPU Settings: Go to Settings → Hardware → GPUs to enable/disable GPU acceleration
  • Restart Jan: Sometimes a fresh start helps

​
Slow Performance

  • Enable GPU acceleration (Settings → Hardware)
  • Use smaller models (3B-7B instead of 13B+)
  • Lower quantization (Q4 instead of Q8)
  • Close other applications to free up RAM

​
Download Issues

  • Check internet connection
  • Verify disk space (10GB+ free recommended)
  • Try different mirror: Some models have multiple download sources
  • Resume failed downloads: Jan automatically resumes interrupted downloads

​
GPU Not Detected

  1. Install NVIDIA drivers (version 470.63.01+)
  2. Install CUDA Toolkit (version 11.7+)
  3. Verify installation: 
    nvidia-smi
nvcc --version
  4. Enable in Jan: Settings → Hardware → GPUs → Toggle ON

Full Troubleshooting Guide

Comprehensive troubleshooting documentation for common issues.

​
Getting Help

If you run into issues:
  1. Check the troubleshooting docs
  2. Search GitHub issues
  3. Ask in Discord #🆘|jan-help channel
  4. Include:
    • Your OS and version
    • Jan version (Help → About)
    • Error messages or logs
    • Steps to reproduce

​
Next Steps

Installation Guide

Platform-specific installation details and GPU setup instructions.

Model Parameters

Learn how to fine-tune model behavior for different use cases.

MCP Integration

Enable agentic capabilities with web search, tools, and integrations.

API Reference

Complete API documentation for building integrations.

Build docs developers (and LLMs) love

Get started for freeTalk to us