Models & Preferences
Choose the Claude model Claw uses for your account.
Model Preference
Claw supports a user-level Claude model preference for its sandboxed runtime.
The current curated options are:
- Claude Opus
- Claude Sonnet
- Claude Haiku
- Claude Sonnet 1M
If you do not choose a model explicitly, Claw defaults to Claude Opus.
How To Think About The Choice
- Opus is the default when you want the strongest all-round reasoning.
- Sonnet is the tradeoff choice when you want a lighter-weight operational model.
- Haiku is the smallest option when you care more about speed than depth.
- Sonnet 1M is the long-context option when the task benefits from larger working context.
What Changes When You Switch
Changing this preference changes the model Claw uses for future sandbox runs on your account. It does not change what Claw is for. The workflows stay the same:
- portfolio questions
- agent reviews
- image and file workflows
- scheduled heartbeats
Practical Guidance
Unless you have a reason to optimize for speed or context window, keep the default.
If Claw becomes part of a heavier reporting or long-context workflow for you, revisit this setting before you change how you prompt.
For the ongoing session model behind this preference, see Memory, Context & Sessions.