Jan

Menlo Research

The open-source ChatGPT that runs on your machine: download a model, chat offline, and keep every token off the network — a local app, not a cloud tab.

Product preview

Overview

Jan is what people install when they decide the cloud is not an option: a desktop ChatGPT clone that pulls GGUF models, runs them on the GPU or CPU you already paid for, and exposes an OpenAI-compatible localhost API so existing tools keep working with the data never leaving the machine.

Compared with LM Studio it is more open-source-project and less polished store app; compared with Ollama it is a full chat UI rather than a daemon you script. Hardware is the real gate — a 7B model is pleasant on a modern laptop, a 70B is not. For private legal, medical or internal docs, “good enough locally” beats “brilliant in someone else’s GPU.”

Pros / Cons

Pros

Offline, no account, no cloud
Open-source desktop app plus local API
Uses the GPU you already own
Drop-in OpenAI base URL for local tools

Cons

Quality capped by local hardware
Model management is still on you
UI polish trails LM Studio

Who it's for

Private chat on air-gapped machines

Local OpenAI-compatible API for tools

Trying open models without a cloud bill

Legal or medical drafts that cannot leave the laptop

Bottom line

ChatGPT that never leaves the laptop — the open-source local app to try before you buy a cloud seat.

Key Features

Local models on macOS, Windows and Linux

Offline chat with no account

OpenAI-compatible local API

Model Hub downloads

Open source, Apache-licensed

Compare Jan with alternatives

Jan vs Hugging Face Jan vs Replicate Jan vs Ollama 4 products · AI Platforms

Frequently asked questions

What is Jan?

Jan is a AI Platforms tool developed by Menlo Research. The company was founded in 2023 and is headquartered in Singapore. The open-source ChatGPT that runs on your machine: download a model, chat offline, and keep every token off the network — a local app, not a cloud tab.