Gemini 3 Is Here—and Google Says It Will Make Search Smarter (www.wired.com)

Google today unveiled Gemini 3, which it bills as its most capable multimodal AI yet—optimized for reasoning, long-horizon planning, multimedia understanding and coding—and is being integrated directly into products like the Gemini app and Search’s AI Overviews starting immediately. Google says Gemini 3 outperforms OpenAI’s GPT-5 on public leaderboards (e.g., LMArena) and shows stronger simulated reasoning (breaking problems into parts) and planning, which improves agents that use tools and the web. In demos the model can generate custom interactive visualizations on the fly, power new developer tools like Antigravity, and support features such as NotebookLM and AI Studio. Google is rolling it out to paid subscribers (around $19.99 and a higher-tier ~$249.90/mo) and plans deeper embeds—potentially into Siri—and into areas from gaming to robotics. The release matters because Google is positioning Gemini 3 not as a standalone chatbot but as a product-integrated platform that enhances core revenue-generating services (Search, Maps, Gmail). Google highlights its data and infrastructure advantages—650M monthly Gemini app users, 2B monthly AI Overview interactions, 13M developers, custom silicon and data centers—as competitive moats for continued model improvement. For practitioners, the key implications are better multimodal understanding, stronger reasoning/planning for agent workflows, and broader real-world deployment opportunities that could shift how retrieval, visualization and tool-using agents are built and scaled.