Google unveiled a multitude of revolutionary updates in the field of artificial intelligence at its annual Google I/O 2025 developers conference, including a new version of Gemini, the Veo 3 video generator, and the Jules AI programming agent.
G. Ostrov
Technology giant Google kicked off its annual Google I/O 2025 developers conference, immediately presenting an impressive array of innovations in artificial intelligence. The first day of the event was marked by announcements of numerous updates to the company's existing AI services and the launch of entirely new projects.
Key Announcements from the First Day of the Conference
Gemini 2.5 Flash — a significant update to Google's language model. The improved version demonstrates superior results in programming, image processing, and working with long context. The model has also become more efficient, consuming 30% fewer tokens. Users can already try the new version for free in AI Studio.
Veo 3 — an advanced video content generator capable of creating videos with sound. This technology will be available to subscribers of the premium AI Ultra plan, which costs $250 per month.
Deep Think for Gemini 2.5 Pro — an innovative mode that, according to developers, allows the model to surpass the performance of competing OpenAI o3 technology.
Gemini Diffusion — a revolutionary approach to language models using a principle characteristic of image generators: instead of sequentially predicting the next token of text, the system generates the entire response from "white noise."
NotebookLM received an important update allowing users to create video reviews based on various information sources, including texts, PDF documents, and websites.
Jules — a full-fledged AI agent for programming in a cloud environment, developed as a direct competitor to OpenAI Codex. Notably, this tool is already available to everyone.
Google Meet now offers a real-time dialogue translation feature, significantly expanding opportunities for international communication and collaboration.
Project Astra represents a substantial update to the voice assistant presented last year. The new version can control devices directly during conversations and retains conversation history in memory.
In addition to the innovations listed, Google announced expanded access to AI Mode in its search system for all users in the United States, as well as new capabilities for developers to manage AI models.
The technology community eagerly anticipates the second day of the Google I/O 2025 conference, as well as the responses of competing companies to the presented innovations.
Official Google I/O conference website: io.google