AI models
Ollama
Run open models on your own machine with one command.
open source self-hostable Ollama Inc. · USThe easiest way to serve Llama, Mistral and friends locally; a US company maintains it.
Website ↗ Source ↗ Plan “AI models” with it
How you can run it
| Way | Where the data lives | Price basis | Verdict |
|---|---|---|---|
| Host it myself | your own server or hardware | free | green — You run it yourself: no data leaves your own infrastructure. |
Facts
| Company / project | Ollama Inc. (US) |
|---|---|
| Licence | MIT |
| Self-hosting footprint | 4096 MB RAM · 2 vCPU · 20 GB — a 7B model on CPU; a GPU makes it pleasant |
| Commonly replaces | OpenAI API |
| Upstream (awesome-selfhosted-data) | 180,079 stars · updated 2026-09-03 · Docker, Python |
| Provenance | curated, verified 2026-09-04 |
Machine-readable: /api/tools/ollama.json. Something wrong? Tell us.