Google I/O

Google I/O featured several interconnected updates to Google’s AI models and tools, with emphasis on agentic systems, multimodal generation, and deeper integration into search and productivity apps. The announcements position Gemini models as production-ready for autonomous tasks rather than purely conversational interfaces, while expanding capabilities in video, coding, and personalized assistance.

Gemini 3.5 Flash serves as the core technical release. Google describes it as its strongest model for coding and long-horizon agentic work, outperforming the prior 3.1 Pro on benchmarks in those areas while running four times faster than competing frontier models. The model supports simultaneous operation of multiple agents on extended tasks and can run autonomously for hours, pausing only at decision points that require human input. It underpins both the Gemini app and AI mode in search, and it was developed alongside the company’s agent platform to reduce latency in parallel workflows. Banks and data science teams are cited as early testers for multi-week automation, though specific results remain undisclosed.