Google I/O adds Gemini agent upgrades and AI search tools

Google unveiled Gemini 3.5 Flash as the new default model for its Gemini app and AI Mode in Search, touting four‑times faster token output and a price that can be as low as one‑third of competing frontier models [2]. The same day, Google introduced Gemini Spark, a general‑purpose AI agent that can act across a user’s connected apps, beginning its beta for Google AI Ultra subscribers at $100 per month [2].

At the I/O keynote, Google also announced “Ask Play,” an AI‑powered overlay that lets users converse with Gemini to find apps in the Play Store, and a suite of developer‑focused upgrades. The Android CLI now runs in a stable 1.0 version with agentic AI integration, letting external coding agents talk directly to the command line and invoke Android Studio capabilities [1]. Google AI Studio, a browser‑based prototyping environment, can now generate full native Android apps in minutes, publish them to internal testing tracks, and export the code to GitHub or ZIP files [1].

The announcements signal a shift from answering queries to taking actions. Sundar Pichai highlighted that Gemini’s monthly active users have risen to over 900 million, with daily requests growing roughly sevenfold and the model APIs processing about 19 billion tokens per minute [3]. By embedding Gemini agents into Android’s 3 billion‑device ecosystem, Google aims to lock developers onto its AI infrastructure, leveraging lower pricing to drive adoption and create recurring subscription revenue beyond ad‑based Search [2].

Google also rolled out Gemini Omni, a model that generates output from any input—including video—paired with a physics‑aware reasoning engine, and began offering AI Overviews to more than 2.5 billion monthly users [3]. These moves position Google’s AI platform as both a developer toolkit and a consumer‑facing service, blurring the line between search, app discovery, and autonomous agents.

The real question is whether the combination of faster, cheaper models and agentic capabilities will translate into sustained developer lock‑in and new revenue streams, or if competitors’ advances—such as Anthropic’s Mythos—will keep the AI battle evenly matched.