Hua Er Jie Jian Wen·2025-05-20 18:57Core Insights - Google emphasizes the significance of its AI assistant Gemini during the 2025 Google I/O developer conference, highlighting the rapid deployment of over ten models and twenty AI features since the last conference [2][4] - The Gemini 2.5 Pro model ranks highly across all categories in the LLM Arena evaluations, and it is the fastest-growing model on the popular AI code editor Cursor [4] - Google introduces Google Beam, an AI-driven video communication platform that enhances video interactions to feel like face-to-face conversations, utilizing advanced 3D technology [6] AI Model Performance - The number of tokens processed by Google's systems has surged from 9.7 trillion to 480 trillion over the past year, marking an increase of nearly 50 times [4] - Gemini's app boasts over 400 million active users monthly, with 1.5 billion users engaging with Google's AI overview in search, indicating a larger user base for generative AI services compared to competitors [4] New Product Launches - Google Beam, an upgraded version of the Starline technology, features six-camera arrays for capturing subjects from multiple angles, achieving near-perfect head tracking with real-time rendering at 60 frames per second [6] - The Gemini 2.5 Pro model is set to enable real-time translation, demonstrated live during the conference, showcasing its capability to switch languages mid-sentence [6] Enhanced Features - Google Meet introduces real-time voice translation, initially supporting English to Spanish, with plans to expand to more languages in the coming weeks [7] - The Gemini 2.5 Flash model operates more efficiently, using 22% fewer tokens while maintaining the same performance level, enhancing overall efficiency for users [7]