Google shifts AI strategy with Gemini 3.5 Flash: agents over chat

Google unveiled Gemini 3.5 Flash at its I/O developer conference on Tuesday, positioning the model as its strongest offering yet for coding and autonomous AI agents. Rather than simply answering questions, the model can independently execute coding pipelines, manage research projects, and, in internal tests, build an operating system from scratch. The launch marks a deliberate pivot from conversational AI to agentic tools—systems that plan, build, and iterate on real work with minimal human input.

DeepMind’s chief technologist, Koray Kavukcuoglu, told reporters that 3.5 Flash combines quality with low latency, outperforming Google’s previous frontier model, 3.1 Pro, on nearly all benchmarks for coding, agentic tasks, and multimodal reasoning. He noted the model is four times faster than other frontier models, with an optimized version achieving 12 times the speed at the same quality. That speed is critical for agentic workflows, where multiple AI agents run concurrently on long-running tasks.

At I/O, Google engineer Varun Mohan demonstrated agents spawning to work on separate components before assembling a full operating system inside Antigravity, the company’s agentic development platform. Kavukcuoglu said Flash 3.5 was co-developed with Antigravity to give agents a native environment to live, work, and execute. Google also released Antigravity 2.0, a standalone desktop app designed for agent-first development.

Early adopters include banks and fintechs automating multi-week workflows, and data science teams extracting insights from complex data. The model can run autonomously for hours, though Tulsee Doshi, Google’s senior director of product, said it will pause for human input at decision points or permission issues. When Google releases its forthcoming 3.5 Pro model, the two will work in tandem—3.5 Pro as orchestrator and planner, Flash as sub-agents.

Gemini 3.5 Flash is now the default model in the Gemini app and AI Mode in Search globally. Google also announced agentic capabilities for Search, letting users create and manage AI agents directly on the platform. The model will power Gemini Spark, a 24/7 personal AI agent for consumers. Google says 3.5 Flash includes strengthened safeguards against cyber and CBRN threats, and is better calibrated to handle sensitive questions. The model is available via Antigravity, the Gemini API, Gemini Enterprise, and the Gemini app.