What probably matters to youSection of the release
Unclear Which models qualify, and what other conditions must hold, is not stated.
What
When you submit a prompt that contains the word ultrathink, and the model you are using qualifies, Claude Code now sets the reasoning effort for that request to xhigh. Reasoning effort is how much thinking the model does before it answers. Claude Code also records that this boost happened.
Why
In practice, including ultrathink in a prompt gets that one request the higher xhigh effort level without changing any setting.
