Picking the right AI for each message.

How Continio picks

Before each message goes anywhere, Continio reads it and decides which AI is best suited. It considers:

• How complicated the task is
• Whether you sound like you want a thinking partner or a quick answer
• Whether there are files involved
• Whether you need fresh information from the web

This happens in milliseconds, on every message. The decision is shown to you. Look under the Continio logo on each reply and you'll see which AI handled it.

Which AI gets used for what

AI	Used for
Claude Anthropic	Default for most things. Coding, reflective questions, writing, planning, anything personal. Picked because it respects continuity rules and tone better than the alternatives.
GPT-4o OpenAI	Anything with an image attached. Complex analysis. When you've selected it manually.
GPT-4o mini OpenAI	Fast, lightweight things. Quick factual questions. Background work like generating chat titles.
Claude Haiku Anthropic	Background work only. Noticing memories. Generating titles. Pattern observation. Not used for replies you read.
Grok xAI	Manual selection only. Different data terms to Anthropic and OpenAI, the chat shows you a note when you pick it.
Gemini Google	Manual selection only. Different data terms, shown when you pick it.

How Continio matches the AI to the kind of message

Alongside picking the AI, Continio also picks how to respond. There are a few different shapes:

• Reflective. Personal or exploratory. Continio leans on what it knows about you, the tone gets warmer.
• Practical. You want a direct answer. Concise, less framing.
• Coding. Code or technical work. Longer replies allowed. File tools switched on.
• Writing. Documents, drafts, structured pieces. Continio can hand you a file at the end.
• Planning. Strategy, decisions, structured thinking.

You don't pick these. They follow from what you've actually written.

Choosing the AI yourself

Tap the AI selector next to the message input. Whatever you pick is used for that message and the rest of the chat. Starting a new chat resets to Auto, unless you pick again.

If you attach an image while you have a non-vision AI selected (Grok, Haiku, GPT-4o mini), Continio switches to GPT-4o just for that message and resumes your selection after.

Auto is usually the right call. Continio's automatic picking is calibrated for how you actually use it. Picking yourself is useful when you specifically want a particular AI's style. It's not because Auto is unreliable.

Web search

Continio has live web search built in. It runs automatically whenever a message would benefit from current information. Prices. News. Links. Recommendations. Anything where an AI's training cut-off would leave it stale.

You don't need to ask. If you ask Continio for links (something to buy, somewhere to find, something to look up), it'll search and give you actual links. Not "here's how you could search for it".

Web search doesn't run when there's an image attached. The AI looking at the image needs the room.

Why this affects energy use

Picking smaller AI for smaller tasks (GPT-4o mini, Haiku) instead of always using the heaviest one is a deliberate choice. It's faster, and it uses meaningfully less energy per message. There's more on this on the Environment page.