I’m really excited about this update. In actual testing, setting the thinking_budget to at least 128 reduced the response time to a quarter,
and now I can even predict the number of tokens in advance to estimate usage costs.
Huge thanks to Google—sincerely appreciate the quick resolution!