On May 20, 2026, at the I/O Developer Conference, Google officially unveiled Gemini 3.5 Flash. This cutting-edge model is capable of processing multimodal inputs, encompassing text, images, videos, audio, and PDFs. It features four distinct levels of thinking intensity, ensuring that the reasoning process is preserved across multi-round dialogues. Designed specifically for multi-agent collaboration, Gemini 3.5 Flash excels in tackling complex tasks, including automated code reviews and batch data analysis. With an impressive output speed exceeding 280 tokens per second, it outperforms its predecessor, the flagship model Gemini 3.1 Pro, in benchmark tests for coding, terminal control, and multimodal comprehension. Furthermore, its cost-effectiveness is notable, with usage costs less than half of those associated with flagship models boasting similar capabilities. Gemini 3.5 Flash has been simultaneously rolled out across multiple products, including the Gemini App, Google Search's AI mode, and Google AI Studio. It has also been designated as the default model for Google Search's AI mode, marking a significant milestone in Google's pursuit of intelligent, lightweight models.