Google's year in review: 8 areas with research breakthroughs in 2025 highlights major advancements across AI, science, and global impact. AI Model Advancements: Gemini 3 models delivered significant improvements in reasoning, multimodality, efficiency, and creativity. Gemini 3 Pro, the most powerful model to date, achieved top rankings on LMArena and excelled on challenging benchmarks like Humanity’s Last Exam and GPQA Diamond. It also set a new state-of-the-art in mathematics with 23.4% on MathArena Apex. Gemini 3 Flash combined Pro-level reasoning with Flash-level speed and cost efficiency, outperforming earlier models at a fraction of the price. AI-Powered Product Integration: AI is transforming core Google products. The Pixel 10 features advanced AI capabilities, while Search now includes agentic features that enable deeper, dynamic interactions. These enhancements allow users to explore ideas, solve problems, and complete tasks more intuitively. Creative Empowerment: Generative AI tools expanded dramatically in 2025. Nano Banana and Nano Banana Pro delivered breakthroughs in image generation and editing. New tools like Flow and Music AI Sandbox supported creative professionals, while upgrades to the Gemini app and the Google Arts & Culture lab introduced AI-driven experiences. Veo 3.1, Imagen 4, and Flow enabled more natural and expressive video, image, audio, and world creation. Scientific and Medical Breakthroughs: AI accelerated progress in life sciences and health. Building on the legacy of AlphaFold, which celebrated its 5-year anniversary, Google AI continues to help researchers understand protein structures. AlphaFold has been used by over 3 million scientists across 190 countries, including over 1 million in low- and middle-income nations. Quantum Computing and Hardware Innovation: Google made strides in quantum computing with projects like Quantum Echoes and celebrated the 2025 Physics Nobel Prize awarded to Googler Michel Devoret and colleagues for foundational quantum research. The introduction of Ironwood, a new TPU designed for inference using AlphaChip, marked a leap in AI hardware efficiency and sustainability. Energy and Infrastructure Efficiency: Google advanced its AI-driven chip design and infrastructure with a focus on reducing environmental impact. Ironwood was developed using AI to optimize performance and energy use, reflecting a broader commitment to sustainable AI development. Climate and Environmental Solutions: AI is helping tackle global challenges. WeatherNext 2 now generates forecasts 8 times faster with up to hourly resolution, supporting cyclone predictions and flood forecasting for over two billion people across 150 countries. These tools are improving disaster preparedness and urban planning. Education and Accessibility: AI is expanding access to knowledge through initiatives like LearnLM and Guided Learning in Gemini. Google Translate now uses the most advanced translation models, delivering more natural, accurate results, and piloting real-time speech-to-speech translation to break down language barriers. These breakthroughs underscore Google’s commitment to advancing AI responsibly, fostering collaboration, and applying technology to solve pressing global challenges.