MOUNTAIN VIEW — Google launched Gemini 3.5 Flash on Tuesday at its annual I/O developer conference, introducing an artificial intelligence model the company says is its strongest yet for coding and autonomous AI agents. The model can independently execute coding pipelines, manage research projects, and, in internal tests, build an operating system entirely from scratch.

Gemini 3.5 Flash is now the default model in the Gemini app and in AI Mode in Search globally. It is available generally through Antigravity, the Gemini API, Gemini Enterprise, the Gemini app, and AI Mode in Search. Agentic capabilities are also coming to Google Search to let users create, customize, and manage AI agents directly on the platform.

"3.5 Flash offers an incredible combination of quality and low latency." DeepMind Chief Technologist Koray Kavukcuoglu said. "It outperforms our latest frontier model, 3.1 Pro, on nearly all the benchmarks, including coding, agentic tasks, and multimodal reasoning."

Kavukcuoglu said the model is four times faster than other frontier models, and that Google developed an optimized version that is 12 times faster at the same quality. Flash 3.5 was co-developed with Antigravity to provide a native environment for agents. "Agents can live, work, and execute in Antigravity as a native environment." he said.

On stage at I/O, Google engineer Varun Mohan demonstrated agents spawning off to work on separate components before building a full operating system inside Antigravity. Google said the model's agentic capabilities are being used by partners including banks and fintechs to automate multi-week workflows, and by data science teams to find insights in complex data environments. The model can run autonomously for multiple hours, pausing to ask for user input when it hits a decision point or permission issue requiring human judgment.

"3.5 Pro becomes your orchestrator, your planner, and then it actually can leverage Flash to be the various sub-agents." Senior Director and Head of Product Tulsee Doshi said. "It really comes down to where you want reasoning power versus where you have tasks that merit good brute force tool use capabilities." she said.

Google said Gemini 3.5 has strengthened cyber and CBRN safeguards and is better calibrated to engage with sensitive questions rather than refuse them outright. Google launched Gemini three years ago with the goal of building a multimodal large language model trained on text, images, audio, and video.