The free, built-in model that will need no key or setup.
A free, built-in model - hosted by RG AI, nothing for you to configure - is in active development. Today, chatting requires bringing your own provider key; the built-in model isn't reliable enough yet to turn on for everyone.
The free model will run on infrastructure RG AI pays for directly, so a daily cap will exist to keep it usable for everyone rather than a few heavy users. Connecting your own key will remove the cap since you're then paying the provider, not RG AI.
Note: In the meantime, bringing your own key to a frontier provider (Claude, ChatGPT, Gemini, Kimi, or DeepSeek) works today and generally gives stronger answers on demanding tasks anyway.