Google I/O 2026: New Gemini app, Flash model, and agentic AI push, here’s everything Google announced

The highly anticipated Google I/O 2026 kicked off on May 19, unveiling a series of groundbreaking announcements centered on artificial intelligence (AI), upgrades to the Gemini app, and a new focus on an 'Agentic era.' CEO Sundar Pichai highlighted the rapid advancements in AI technology, citing a significant evolution in models, infrastructure, and applications that touch billions of lives globally. This transformative phase, which Pichai referred to as 'hyper progress,' marks a shift where AI transitions from experimental stages to an integral part of daily life. Among the standout revelations was the remarkable increase in AI usage. Google now processes over 3.2 quadrillion tokens monthly—an astonishing rise from the 480 trillion tokens recorded at last year's event. This surge reflects the growing integration of AI features across various products. Google's AI Overviews have now attracted 2.5 billion monthly active users, while the AI Mode boasts 1 billion. The Gemini app alone has seen its user base double, soaring from 400 million to 900 million within a year, alongside the Nano Banana tool, which has facilitated the creation of over 50 billion images globally. In response to the burgeoning demand for AI, Google has pledged a substantial investment of $180 to $190 billion in infrastructure, primarily directed towards developing custom Tensor Processing Units (TPUs). At the event, two new eighth-generation TPU architectures were introduced: TPU 8t, designed for large-scale AI training, and TPU 8i, optimized for inference workloads. Another highlight was the launch of Gemini Omni, a new family of multimodal models tailored for AI video generation. This innovative model enables users to create outputs across various formats and supports conversational editing, allowing for dynamic adjustments using voice commands. The initial rollout features Gemini Omni Flash, aimed specifically at video generation, with plans for future expansions into image and text generation. Additionally, Google unveiled its flagship AI model, Gemini 3.5 Flash, which promises enhanced performance in coding, reasoning, and multi-model comprehension, delivering four times the speed of its predecessors. The new version of the agentic coding application, Antigravity 2.0, allows users to operate multiple specialized AI agents, create custom workflows, and automate tasks seamlessly in the background. The Gemini app itself received a significant makeover, featuring a refreshed interface with smooth animations, vibrant colors, and haptic feedback, collectively termed 'Neural Expressive design.' Enhancements include upgraded performance powered by Gemini 3.5 Flash, AI video generation, and new agent features. The introduction of Gemini Spark, a personal AI agent capable of operating continuously on behalf of users, further exemplifies Google’s commitment to AI integration within its Workspace tools. This agent can execute tasks autonomously while maintaining contextual awareness. In addition to these developments, Google previewed audio-based smart glasses, set to launch later this year, in collaboration with Warby Parker and Gentle Monster. These glasses will utilize Google's Gemini AI model for enhanced functionality. The event also highlighted Gemini for Science, a suite of AI tools designed to aid researchers, and introduced new AI features for YouTube, such as Ask YouTube for enhanced video searching and Docs Live for streamlined documentation generation. Overall, this year's I/O showcased Google’s ambitious vision for an agentic AI ecosystem, emphasizing intelligent systems designed to assist users, automate tasks, and seamlessly integrate across diverse platforms and devices.