Best self-hosted ChatGPT alternatives
Short answer: The most common self-hosted ChatGPT alternative is Ollama (to run models) plus Open WebUI (the chat interface); LibreChat suits multi-provider setups, AnythingLLM chatting with documents, and Jan a desktop app with no server.
Last updated: · No tax figures on this site, so no GOV.UK check applies; tool details come from each project's public GitHub repository · Markdown version
A self-hosted ChatGPT setup needs two parts: a model runtime and a chat interface.
| Tool | Type | Licence | Best for |
|---|---|---|---|
| Open WebUI | Chat UI | BSD-3 | ChatGPT-style UI for Ollama and APIs |
| LibreChat | Chat UI | MIT | Multi-provider chat with agents |
| AnythingLLM | RAG app | MIT | Chat with your documents |
| Jan | Desktop app | Apache-2.0 | Offline, no server |
| Ollama | Runtime | MIT | One-command local models |
What does a typical self-hosted stack look like?
Laptop: Jan, or Ollama + Open WebUI. Small team on a VPS: Ollama or vLLM + LibreChat or Open WebUI, with Langfuse for observability. See Ollama vs llama.cpp vs vLLM.
Frequently asked questions
What is the easiest self-hosted ChatGPT alternative?
For a single machine, Ollama to run models plus Open WebUI as the chat interface is the most common setup. Jan is a desktop app if you do not want Docker.
Do I need a GPU?
Not for small quantised models: llama.cpp and Ollama can run them on CPU, but slowly. For larger models or several users, a GPU makes a big difference.
Can I use cloud APIs from the same interface?
Yes. Open WebUI, LibreChat and AnythingLLM can connect to OpenAI-compatible APIs as well as local models.