Google launched Gemini 3.5 Flash on Tuesday at its annual I/O developer conference, positioning the model as its strongest system yet for coding and autonomous AI agents.
The company says the model can execute coding pipelines, manage research projects, and, in internal tests, build an operating system from scratch. The launch shows Google’s shift from AI as a conversational tool to AI as an agentic tool that can plan, build, and revise work with less human input.
4x Faster Than Rivals
Koray Kavukcuoglu, DeepMind’s chief technologist, told reporters that Gemini 3.5 Flash offers strong quality with low latency.
He said the model outperforms Gemini 3.1 Pro on nearly all benchmarks, including coding, agentic tasks, and multimodal reasoning. He also said Gemini 3.5 Flash is four times faster than other frontier models, while an optimized version of Flash is 12 times faster with the same quality.
That speed is central to Google’s agentic AI plans. The company designed the model for workflows where multiple AI agents can run at the same time on long-running tasks.
Antigravity 2.0 for Developers
At I/O, Google also released Antigravity 2.0, a standalone desktop app built around agent-first development.
Google engineer Varun Mohan demonstrated AI agents working on separate parts of a project before coming together to build a full operating system inside Antigravity. Kavukcuoglu said Gemini 3.5 Flash was co-developed with Antigravity so agents could have a native environment to work and execute tasks.
Google says Antigravity 2.0 lets developers orchestrate multiple agents in parallel. It also includes dynamic subagents, scheduled background tasks, and integrations with Google AI Studio, Android, and Firebase.
Managed Agents Come to Gemini API
Google is also adding Managed Agents to the Gemini API.
With a single API call, developers can create an agent that reasons, uses tools, and executes code inside an isolated Linux environment. These agents use the Antigravity agent harness, which is built on Gemini 3.5 Flash and is available through the Interactions API and Google AI Studio.
Google says the model is already showing results beyond demos. According to the company, banks and fintech firms are using its agentic capabilities to automate multi-week workflows, while data science teams are using it to find insights in complex data environments.
Human Input Still Needed At Key Points
Gemini 3.5 Flash can operate autonomously for multiple hours.
However, Tulsee Doshi, Google’s senior director and head of product, said the model may pause and ask for user input when it reaches a decision point or a permission issue that requires human judgment.
Google also plans to release Gemini 3.5 Pro later. Doshi told TechCrunch that 3.5 Pro will work as an orchestrator and planner, while Flash can act as the sub-agent for tasks that need fast tool use.
Search Agents and Personal Intelligence
This summer, Google will introduce information agents in Search for Google AI Pro and Ultra subscribers.
These agents will work in the background around the clock and monitor the web, news sites, social posts, and fresh data from finance, shopping, and sports. For example, users looking for an apartment can give Google their exact requirements and receive updates when matching listings appear.
Google is also expanding Personal Intelligence in AI Mode to nearly 200 countries and territories across 98 languages. The feature does not require a subscription and can connect with Gmail and Google Photos, with Google Calendar support coming soon.
Gemini Spark and Safety Concerns
Gemini 3.5 Flash will also power Gemini Spark, Google’s personal AI agent designed to run 24/7 and help users manage their digital lives.
The wider release of autonomous AI tools comes as Google faces legal scrutiny over Gemini. Reuters reported in March that the family of Jonathan Gavalas sued Google, alleging that Gemini drove him into paranoia and later suicide. Google said Gemini is designed not to encourage real-world violence or suggest self-harm and that it continues to improve safeguards.
Google says Gemini 3.5 includes stronger cyber and CBRN safeguards and is better calibrated to handle sensitive questions instead of refusing them outright.
Gemini 3.5 Flash is generally available through Antigravity, the Gemini API, Gemini Enterprise, the Gemini app, and AI Mode in Search.
