Google launches Gemini 3.5 Flash — an agentic AI that codes, manages projects and builds operating systems

5/20/2026, 08:49 AMМария Фадеева

Google has unveiled Gemini 3.5 Flash, the first model in its new Gemini 3.5 family, positioning it as the most capable AI system available for coding and agentic tasks. The model can write code autonomously, manage research projects, and build operating systems from scratch.

The defining shift in Gemini 3.5 Flash is the move from conversational AI to agentic AI: rather than simply responding to queries, the system is designed to execute complex, multi-step tasks across extended workflows. Google says the model generates tokens four times faster than comparable advanced models. For higher-stakes actions — such as sending emails or handling financial transactions — the agent will request explicit user approval before proceeding. A Gemini 3.5 Pro version is expected to follow next month.

Alongside the model itself, Google announced a broader ecosystem of products.

Antigravity 2.0 serves as a central platform for coordinating multiple AI agents running in parallel. It supports dynamic sub-agents for different workflows, scheduled background automation, and deep integration with Google AI Studio, Android, and Firebase. The release also includes Antigravity CLI for spinning up agents without a graphical interface, Antigravity SDK for programmatic access to the same agent tooling used in Google's own products, and an Antigravity integration within the Gemini Enterprise Agent Platform for Google Cloud customers.

Gemini Spark is a personal AI assistant built on Gemini 3.5, designed for everyday tasks with native access to Google's service ecosystem.

Gemini Omni is a multimodal model combining text, image, and video generation. Users can produce high-quality video content with cinematic zoom effects and prompt-driven background replacement.

Google Pics is a new image creation and editing tool capable of generating everything from event posters to infographics, with automatic object detection and relationship analysis.

For developers, Google introduced managed agents in the Gemini API: a single API call can now launch an agent capable of reasoning, using tools, and executing code in an isolated Linux environment, all built on the Antigravity platform.

A new Google AI Ultra subscription tier has been launched at $100 per month, offering five times the Antigravity usage limit of the AI Pro plan. Google also previewed a pair of smart glasses, with a design collection slated for release this autumn.

Popular news