- Added new model `gemini-3.5-flash` to the registry with enhanced intelligence and speed capabilities. - Supports extended thinking levels (`minimal`, `low`, `medium`, `high`) and dynamic adjustments. - Expanded generation methods, including content creation and token counting.