Google DeepMind Unveils Gemini 2.5, Imagen 4, and Gemma 3n, Showcasing Broad AI Advancements to Over 1,000 Developers at I/O Connect Berlin

Berlin – Google DeepMind recently showcased significant advancements in its artificial intelligence portfolio during the I/O Connect Berlin event, drawing over a thousand developers, founders, and builders. Patrick Loeber, who led a key session on the main stage, highlighted the expanded capabilities of the Gemini API, including the new Gemini 2.5 model family and advanced features. The event underscored Google's commitment to empowering the global developer community with cutting-edge AI tools.

During the session, Loeber demonstrated the versatility of the Gemini API, emphasizing the powerful Gemini 2.5 models, which now include the advanced "Deep Think" mode for enhanced reasoning in Gemini 2.5 Pro. He showcased features such as text-to-speech capabilities, which support multi-speaker output and over 24 languages, and the Live API, enabling more natural and expressive voice interactions. These updates aim to provide developers with greater control over AI responses and integrate seamlessly into various applications.

A core focus of the presentation was the evolution of agentic capabilities, allowing AI systems to perform complex tasks autonomously. This includes Project Mariner, a browser-based agentic AI capable of handling multiple web-based tasks simultaneously, and tools like Jules for coding assistance and Stitch for app design. These agentic advancements signify a move towards more proactive and integrated AI assistance for developers and end-users alike.

Beyond the Gemini API, Google also announced the launch of several other key AI products during the same week. This includes the Gemini CLI (Command Line Interface), designed to streamline developer workflows, and Imagen 4, Google's latest image generation model offering higher quality, photorealism, and improved text accuracy. Additionally, Gemma 3n was introduced as a new open-model family, bringing enhanced multimodal input and efficient on-device AI capabilities for mobile and wearable applications.

Loeber expressed enthusiasm for the strong turnout at I/O Connect Berlin, stating in a tweet, "I loved seeing over a thousand developers, founders, and builders coming together from all over the world." He added, "There's never been a better time to be a builder." The event served as a crucial platform for networking and collaboration, reinforcing Google DeepMind's vision for an accessible and innovative AI ecosystem.