Models
Different jobs want different brains. Quick lookups want speed. Hard reasoning wants depth. Research wants something that can read the web all day.
By default your Chief of Staff picks for you. You don't have to think about it. But the picker is right there if you want to override it.
What it does for you
Auto by default:
- "Optimized" mode reads your message and picks the right model for the task
- Quick pulls and short replies use the fast model — cheap and instant
- Drafting, copy, and reasoning use the smart model — better answers
- Long research runs use the research model — reads the web in depth
- You stay in the chat — no tabs, no settings to learn
Override when you want to:
- One-tap model picker in the composer's bottom toolbar
- Models are grouped by capability — Fast, Balanced, Premium, Frontier
- Relative cost shown as dollar signs ($ to $$$$$)
- Your pick sticks for the rest of that conversation
- Set your own Default Model in preferences and every new chat starts on it
Built into every chat:
- Same memory and skills regardless of which model is running
- Same Feed, same documents, same connected apps
- Switch mid-conversation without losing context
How you use it
- Open the chat. You start on Optimized — or on your own Default Model if you've set one.
- Ask anything. Your CoS routes to the right model behind the scenes.
- To override, tap the model button in the toolbar under the message box.
- Pick from the groups — Fast, Balanced, Premium, or Frontier.
- Send. Your choice stays pinned for this conversation.
If you don't have a strong opinion, leave it on Optimized. That's what it's there for.
If you only do one thing today: leave it on Optimized for a week. If you ever feel you needed something smarter, that's when to start playing with the picker.