Google retooled nearly its entire product line around Gemini at I/O 2024, announcing that AI Overviews are rolling out to all U.S. Search users this week. The feature, powered by a custom Gemini model, brings multi-step reasoning and planning to search results. Google also opened a waitlist for developers to access Gemini 1.5 Pro with a 2-million-token context window.

Project Astra, described by TechRadar as an AI agent for everyday life, was demonstrated on a Pixel phone, parsing live video and audio to answer questions in real time. Separately, Google introduced Veo, a video generation model producing 1080p clips that can go beyond a minute, and Imagen 3 for photorealistic image generation with improved text rendering.

Gemini 1.5 Flash, a lighter model optimized for speed, entered public preview alongside significant improvements to Gemini 1.5 Pro. On Android, Gemini Nano gains multimodal capabilities later this year, starting with Pixel, enabling on-device understanding of sights, sounds, and speech. Gemini 1.5 Pro is now available in the side panel in Gmail, Docs, Drive, Slides and Sheets via Workspace Labs, and it is rolling out to Gemini for Workspace customers and Google One AI Premium subscribers next month. A new Ask Photos feature will use Gemini to search personal photo libraries.