Settings

Choose which models the assistant offers, and which one new chats start with.

Assistant models

Sonnet 4.6Balanced$3/$15 per MtokBest balance of speed, intelligence, and cost.

Haiku 4.5Fast & cheap$1/$5 per MtokFastest and cheapest. Great for quick lookups and capture.

Opus 4.8Most capable$5/$25 per MtokDeepest reasoning. Pricier; best for hard questions.

Opus 4.7Capable$5/$25 per MtokPrevious-generation Opus.

Stored on this device. You can still switch models mid-chat from the picker.

Custom models & API keys

Planned

Today the assistant uses Anthropic models through a server-side key that never reaches the browser. Adding other providers, or your own custom models, means storing each provider key securely on the server (encrypted), with a small admin form here. That is the next build.