Ollama

Run open models like Llama, Qwen and Mistral on your own server.

Install on your server Website

An API for open language models, on your hardware: no data leaves the server. Pull a model with the API, then point Open WebUI or your own app at this address. A model needs memory: about 8GB of RAM for a 7B model.

What moin adds

HTTPS on your domain, checked every minute, with an alert when it goes down.
New releases the hour they ship. Update in one click, with a snapshot first.
Secrets made for you, encrypted, never in a file.

Versions

0.40.14 released 2d ago
0.40.13 released 7d ago
0.40.12 released 8d ago
0.40.11 released 12d ago
0.40.10 released 13d ago
0.40.9 released 13d ago
0.40.8 released 14d ago
0.40.7 released 15d ago
0.40.6 released 19d ago
0.40.5 released 21d ago
0.40.4 released 22d ago
0.40.3 released 25d ago

and 145 older, every one installable

License: MIT
Image: ollama/ollama
Source: github.com/ollama/ollama