Back to glossary

Term

Jan

Jan is an open-source desktop application that lets you run large language models locally and fully offline on your own computer. It uses the llama.cpp engine and is positioned as a privacy-friendly ChatGPT alternative.

Jan — explained in detail

Jan is an open-source desktop application (for Windows, macOS and Linux) positioned as a privacy-friendly alternative to ChatGPT. Its core feature is the ability to run large language models (LLMs) locally on your own hardware — fully offline after the initial setup and model download, meaning your inputs never leave the device.

Technically Jan builds on the llama.cpp engine, which runs models in the GGUF format efficiently on CPU and GPU. Through a graphical interface you can download models (such as Llama, Mistral, Qwen, Gemma) from sources like Hugging Face and use them in chat. Jan is open source under the AGPLv3 license and is developed by the Menlo Research team (GitHub organisation janhq).

Feature scope

  • Local operation + optional cloud: By default models run locally, but Jan can also connect to cloud providers (OpenAI, Anthropic/Claude, Mistral and others) if you want to use their models via an API key.
  • OpenAI-compatible local API: Jan exposes a local server (typically at localhost:1337) that mimics the OpenAI API. Other programs can address the local model as if it were a cloud service.
  • MCP support: Via the Model Context Protocol, tools can be connected, enabling agentic workflows.

Context

Jan belongs to the category of tools for local LLMs. Its main benefit is privacy and independence: sensitive data stays on your own device, there are no ongoing API costs, and the tool works without an internet connection. In return, your own hardware limits model size and speed.

Example / Practical use

A typical use case: you install Jan, download a suitable model in GGUF format through the interface, and can then chat with the model offline — for example to draft text or summarise confidential documents without sending them to an external service. If you want your own application to use the model, you point it at Jan’s local OpenAI-compatible API.

  • Jan vs. Ollama: Both run LLMs locally. Ollama is primarily a command-line/server tool often used as a backend; Jan is a full desktop app with a chat UI. Both offer an OpenAI-compatible API.
  • Jan vs. LM Studio: Both are graphical desktop apps for local models. Key difference: Jan is open source (AGPLv3), LM Studio is proprietary (free to use).
  • Jan vs. llama.cpp: llama.cpp is the underlying inference library/engine; Jan is an application on top that makes llama.cpp usable together with model management and a UI.
  • Jan vs. ChatGPT: ChatGPT is a hosted cloud service running the provider’s models; Jan is a local application in which you run models of your own choice, mostly open source.

Entdecke mehr

Themenuebersicht