Google's year in review: eight research breakthroughs in 2025

2025 brought a run of notable AI research breakthroughs across eight areas, led by models such as Gemini 3 and Gemma 3. Improvements in reasoning, multimodality and efficiency translated into new products and features across the company’s portfolio, with an emphasis on responsible development and collaboration as research addressed global challenges.

Progress on world-class models continued through releases including Gemini 2.5, Gemini 3 and Gemini 3 Flash. Gemini 3 Pro pushed multimodal reasoning and mathematics capabilities, while Gemini 3 Flash combined Pro-grade reasoning with lower latency and cost; the Gemma family remained lightweight and open, gaining multimodal features, larger context windows, multilingual support and efficiency gains.

AI shifted from a tool to an active utility across products, powering agentic developer systems such as Google Antigravity and boosting coding with Gemini 3, while adding features to Pixel 10, Search’s AI Mode, the Gemini app and NotebookLM.

gemini 3, gemma 3, gemini 2.5, gemini flash, gemini pro, multimodality, multimodal reasoning, efficiency gains, google antigravity, pixel 10