Google’s flagship Gemini 4 reasoning model — text, image, and video with deep reasoning and 1M-token context.