What will you mostly use this for? This just picks a starting default model — you can change it anytime from the Model dropdown at the top, or start a new chat in the other category from the sidebar.
This downloads once into your browser and then runs entirely on your own GPU — nothing you type leaves your machine. Bigger models are more capable but take longer to download and need more GPU memory; you can always switch later from the Model dropdown at the top.
This is a light coding assistant. If your browser supports WebGPU and you allow it, the model downloads once and runs entirely on your GPU — nothing you type leaves your machine. Otherwise it falls back to this server's CPU-hosted model. Chats are saved in this browser (not on the server) so you can pick up where you left off, and each one belongs to a category — start a new chat from the sidebar to switch.