How to run your own ChatGPT on a VPS

You can have your own ChatGPT-style interface on a VPS, in two ways: with models that run locally on the server, or with an interface that uses your own API keys. In both cases, the conversations stay on your server.

Option 1: local models with Ollama

Ollama + Open WebUI runs open-source models (Llama, Mistral and others) directly on the server, with no per-message cost and without data leaving the server. You need more RAM (at least 8 GB recommended for small models). It installs with one click from the app catalog.

Option 2: interface with your own API keys

LibreChat gives you a ChatGPT-style interface for your team, but uses your own API keys from OpenAI, Anthropic or other providers. Ideal if you want the large commercial models, but with a single shared account and the history on your server, not at the provider.

Which plan you need

• Interface with API keys (LibreChat): a small plan is enough; the heavy lifting is done by the AI provider\'s servers.
• Local models (Ollama): pick a plan with more RAM; the larger the model, the more memory you need.

Why self-hosted

Running the AI on your server keeps your data and keys with you, you have no usage limits imposed by a subscription, and the cost is predictable, billed hourly for as long as the server runs. You can even stop the server when you are not using it, so you do not pay for nothing.