Google used its I/O 2026 keynote to showcase where its AI ambitions are heading next, and this year’s event was heavily focused on making Gemini faster, smarter, and far more capable across everyday tasks.
The keynote began with Sundar Pichai introducing a series of new AI-powered tools and upgrades designed for consumers, developers, and enterprise users. One of the early highlights was Docs Live, a new feature aimed at improving real-time collaboration inside Google Docs using AI assistance.
But the biggest moment of the event arrived later when Demis Hassabis took the stage and introduced Gemini Omni, a new multimodal AI model that pushes Google deeper into AI-generated video creation.
Gemini 3.5 Series Focuses On Speed And Agentic AI
Alongside Gemini Omni, Google also announced the Gemini 3.5 series models, which succeed the Gemini 3.1 lineup. The company says the new models bring major improvements in coding, reasoning, and agentic capabilities, an area where AI systems can independently perform complex tasks using multiple tools and workflows.
The first model rolling out globally is Gemini 3.5 Flash. It is now available through the Gemini app, AI Mode in Search, AI Studio, Android Studio, and Google’s Antigravity developer platform.
Google claims Gemini 3.5 Flash delivers flagship-level AI performance while operating significantly faster and at a lower cost compared to rival frontier models. According to the company’s internal benchmark testing, the model outperformed competitors such as Anthropic’s Claude Sonnet 4.6, Claude Opus 4.7, and OpenAI GPT-5.5 across multiple evaluations related to agentic workflows, multimodal understanding, financial analysis, and long-context processing.
The company also highlighted the model’s output speed, saying Gemini 3.5 Flash can generate responses at 289 tokens per second, which Google says is faster than competing flagship AI systems.
Google Demonstrated AI Agents Building An Operating System
One of the more eye-catching demonstrations during the keynote focused on Gemini 3.5 Flash’s agentic capabilities.
Google revealed that the model was able to use the Antigravity platform to generate a fully functional operating system in roughly 12 hours. According to the company, the process involved 93 AI agents running in parallel and cost under $1,000 in API usage.
The demonstration appeared to underline Google’s growing focus on AI agents that can independently plan, coordinate, and execute complex tasks instead of simply responding to prompts.
Gemini Omni Is Google’s Most Ambitious Video AI Yet
While Gemini 3.5 Flash focused on performance and productivity, Gemini Omni was clearly designed to showcase Google’s next step in generative media.
Google describes Gemini Omni as its first video generation model capable of handling fully multimodal prompts. Users can reportedly combine text, images, audio, and video within a single prompt to create or edit content.
At the moment, the Omni Flash model is primarily being used for AI video generation. However, Google also demonstrated conversational video editing features, allowing users to modify specific elements inside videos, including backgrounds, objects, and characters, simply by describing the changes naturally.
ALSO READ: Apple WWDC 2026 Dates Announced: iOS 27, Smarter Siri And Foldable iPhone Features Expected
The feature feels like a major expansion of Google’s AI creative tools and could eventually become a serious competitor to other emerging AI video platforms.
Rollout Details
Gemini Omni Flash is now rolling out globally to Google AI Plus, Pro, and Ultra subscribers through the Gemini app and Google Flow.
Google also confirmed that the technology is beginning to arrive on YouTube Shorts and the YouTube Create app starting this week, with some features being offered to users for free.