Jan
Description
An open-source alternative to ChatGPT that can run entirely offline on your own computer. In the app you download open models (Llama, Gemma, Qwen, gpt-oss and others from Hugging Face) and chat with them, or plug in online models from OpenAI, Anthropic, Mistral and other providers with your own account. It starts a local server compatible with the OpenAI API (localhost:1337) so other programs can connect, and supports MCP for agent tasks. It is available for Windows 10 and later, macOS 13.6 and later, and Linux (.deb and AppImage), with GPU acceleration on NVIDIA, AMD and Intel Arc cards, and CPU-only builds exist, so a graphics card is not required. The code is under the Apache 2.0 license. The app has no model of its own. It is made by Menlo Research (checked 25 October 2026).
Pros and cons
Pros
- Free and open source (Apache 2.0)
- Windows, macOS and Linux
- Local models and online models in one chat
- Local OpenAI-compatible server and MCP support
Cons
- Large models need a lot of memory; the README suggests 16 GB of RAM for 7B models
- Online models need your own accounts with their providers
Key facts
- Free access
- Yes Free and open source; online models connected through your own accounts are billed by their providers
- Developer
- Menlo Research
- Checked on
- Oct 25, 2026