That has been my experience. Primarily because it is allowed to expend far more test-time tokens than Gemini 3.0 Pro to solve the same prompt.