From what I can tell, their 'local' mode still can't use your own LLM model, just host it yourself, and still charged per request despite using my own hardware

From what I can tell, their 'local' mode still can't use your own LLM model, just host it yourself, and still charged per request despite using my own hardware to run? Very confusing.