What Did Google Announce at Google I/O 2026?

TL;DR

Google’s I/O 2026 announcements center on faster Gemini models, autonomous agents, AI-powered Search, and tools that turn ideas into documents, images, videos, and actions. Gemini 3.5 Flash is available across Google products and APIs, while Gemini Spark, agent-based Search, intelligent shopping systems, Google Pix, an upgraded Flow, and audio glasses extend AI into everyday tasks and creative work.

Transcript

Hello Shoreline and hello to everyone watching from around the world. Excited to be back for this year's I/O. I think it's been an incredible year. We are taking a differentiated, full stack approach to AI innovation to understand the scale at which people are adopting AI. There's another great proxy tokens. The fundamental units of data are model... Read More

Key Insights

• Google’s AI adoption is expanding rapidly across its products. The company reported 13 products with more than a billion users each, including five with more than three billion, while monthly Gemini app usage rose from 400 million to more than 900 million within a year.
• AI Search is becoming the foundation of the Google Search experience. AI Overviews has more than 2.5 billion monthly users, AI Mode has surpassed 1 billion monthly users, and planned agents can monitor information, produce synthesized updates, and build reusable planners, trackers, dashboards, and tools.
• Google’s training infrastructure distributes workloads across multiple sites. JAX and Pathways enable training across TPUs globally, giving Google the ability to create its largest training cluster while supporting the scale required for increasingly capable models and the rapid growth in tokens processed across its services.
• Gemini 3.5 Flash is designed to combine strong intelligence with high-speed action. Google demonstrated an upcoming Flash model generating nearly 1,500 tokens per second and said 3.5 Flash is 12 times faster in Anti-Gravity, with availability beginning immediately across products and APIs.
• Anti-Gravity supports coordinated teams of AI subagents for complex software projects. In one demonstration, 93 subagents worked for 12 hours, made more than 15,000 model requests, processed 2.6 billion tokens, and produced the core of a functioning operating system from an initially empty project.
• Gemini Spark is a personal AI agent that performs long-running tasks under the user’s direction. It operates continuously on dedicated Google Cloud virtual machines, continues after a laptop is closed, connects with Google tools, and is planned to support third-party tools through MCP.
• Google’s agent commerce system has three stated building blocks: the Universal Commerce Protocol, the Agent Payments Protocol, and a universal cart. The cart can reason about products already selected, identify incompatibilities such as mismatched processor and motherboard sockets, and recommend a suitable alternative.
• Google’s creative AI tools support editable images and parallel video generation. Google Pix can remove or resize elements, edit or add text, and translate content, while the upgraded Google Flow can analyze an image, identify useful angles, and transform one image into 16 distinct videos.

Questions & Answers

Q: What were the main Google I/O 2026 AI announcements?

Google announced Gemini 3.5 Flash, previewed Gemini 3.5 Pro, introduced the multimodal Gemini Omni model, and presented Gemini Spark as a personal agent for long-running tasks. It also expanded agent capabilities in Search and shopping, added the Daily Brief, introduced Google Pix for image creation and editing, upgraded Google Flow, and announced audio glasses scheduled to arrive in the fall.

Q: How widely are Google’s AI products being used?

Google reported that AI Overviews has more than 2.5 billion monthly users and AI Mode has surpassed 1 billion monthly users. The Gemini app grew from 400 million monthly active users at the previous I/O to more than 900 million. Across the company, 13 products have more than a billion users each, and five exceed three billion users.

Q: What is Gemini 3.5 Flash and how fast is it?

Gemini 3.5 Flash is the first model in a new series that Google describes as combining frontier intelligence with action. The company said it performs better than 3.1 Pro Flash across almost all benchmarks and is much faster. It has been optimized to run 12 times faster in Anti-Gravity and is available across Google products and APIs.

Q: How did Google use AI agents to build an operating system?

Google used Gemini 3.5 Flash and its Anti-Gravity agent development platform to build the core of a functioning operating system from an empty project. During the 12-hour task, 93 subagents worked in parallel, made more than 15,000 model requests, and processed 2.6 billion tokens. Additional agent work created missing video and keyboard drivers so the system could run Doom.

Q: What is Gemini Spark and what tasks can it perform?

Gemini Spark is a personal AI agent that acts on a user’s behalf and under the user’s direction. It runs continuously on dedicated Google Cloud virtual machines, so tasks can continue after the laptop is closed. Demonstrations included tracking event responses, updating a Google Sheet, emailing people who had not replied, and creating a school-year planning document.

Q: How will AI agents change Google Search?

Google plans to let users create and manage multiple AI agents directly within Search. Information agents can run continuously in the background, monitor real-time data, and send synthesized updates when relevant changes occur. Search can also create stateful experiences such as tools, trackers, dashboards, and weekend planners connected securely to services including Gmail, Photos, and Calendar.

Q: How does Google’s intelligent shopping cart work?

Google’s universal cart is designed to reason about products as soon as they are added and continue working in the background. In the demonstration, it recognized that a selected processor required a different motherboard socket than the chosen motherboard provided. The cart flagged the compatibility problem and suggested an alternative before the user completed the purchase.

Q: What creative AI tools did Google introduce or update?

Google introduced Google Pix in Workspace as an image creation and editing tool with direct creative controls. Users can remove elements, resize objects, add or edit text, and translate content. Google also upgraded Flow so an agent can perform multiple actions simultaneously, analyze an image, determine compelling visual angles, and turn a single source image into 16 unique videos.

Summary & Key Takeaways

• Google presented a full-stack AI strategy spanning infrastructure, models, applications, and consumer services. Adoption has grown across Search and the Gemini app, while JAX, Pathways, distributed TPU infrastructure, and faster model generation support increasingly demanding workloads. The company positioned Gemini as a central reason people use its products more frequently.

• Gemini 3.5 Flash combines frontier-level capability with greater speed and is available across Google products and APIs. Google demonstrated its agentic potential through Anti-Gravity, where 93 collaborating subagents processed 2.6 billion tokens and made over 15,000 model requests while building the core of an operating system from an empty project.

• Google expanded AI from answering questions to completing ongoing tasks. Gemini Spark can work in the background across connected tools, Search can create agents and stateful experiences, and commerce features can manage carts and detect incompatibilities. Google also introduced creative tools for image editing, parallel video generation, personalized daily briefings, and upcoming audio glasses.