You can run very powerful llm locally, albeit slowly if you lack GPU power.

Check out Dolphin models you can run on LM Studio