Jan: Open-Source ChatGPT Alternative for Local AI

Installation

Jan is available for macOS, Windows, and Linux. Download from jan.ai or use a package manager. The app is ~200 MB; models are stored separately.

# macOS (Homebrew)
brew install --cask jan

# Windows (winget)
winget install Homebrew.jan

# Linux (AppImage — download from jan.ai/download)
chmod +x jan-linux-x86_64-*.AppImage
./jan-linux-x86_64-*.AppImage

# Jan stores data at:
# macOS: ~/Library/Application Support/jan/
# Windows: %APPDATA%\jan\
# Linux: ~/.config/jan/

# Models are in: <data_dir>/models/

Jan Hub: Model Discovery & Download

Jan Hub is the built-in model registry with curated, pre-configured GGUF models. Each model listing shows size, required RAM, speed estimate, and quality notes. One click downloads and pre-configures the model with recommended settings.

# Jan Hub featured models (2026)
# Model                    Size    RAM     Use Case
# ─────────────────────────────────────────────────
# Llama 3.2 3B (Q4_K_M)   2.0 GB  ~4 GB   Fast chat, mobile-class hardware
# Llama 3.2 7B (Q4_K_M)   4.1 GB  ~6 GB   General purpose (recommended start)
# Mistral 7B v0.3          4.1 GB  ~6 GB   Instruction following, coding
# Qwen2.5 7B Instruct      4.4 GB  ~7 GB   Multilingual, reasoning
# Gemma 3 9B (Q4_K_M)     5.4 GB  ~8 GB   Google's efficient model
# Phi-4 14B (Q4_K_M)      8.4 GB  ~12 GB  Strong reasoning, smaller footprint
# DeepSeek-R1 7B           4.7 GB  ~7 GB   Chain-of-thought reasoning

# To import a custom GGUF from HuggingFace:
# Settings > My Models > Import Model > paste HuggingFace URL

Using the Jan Local API Server

Jan's API server implements the OpenAI specification on port 1337. Enable it in Settings > API Server > Start Server. The server auto-starts with the selected model loaded.

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

# Chat completion
curl http://localhost:1337/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "llama3.2-7b-instruct",
    "messages": [{"role": "user", "content": "Summarize quantum computing"}],
    "stream": false,
    "max_tokens": 512
  }'

# Using Python openai SDK
from openai import OpenAI
client = OpenAI(base_url="http://localhost:1337/v1", api_key="jan")
completion = client.chat.completions.create(
    model="llama3.2-7b-instruct",
    messages=[{"role": "user", "content": "Write unit tests for this function..."}]
)
print(completion.choices[0].message.content)

Jan Extensions Ecosystem

Extensions are installed from Settings > Extensions. Key extensions available in 2026:

Jan vs Ollama vs LM Studio