What's New in AI

Developers can now upload files for use in chat and search uploaded knowledge bases with Collections Search. Conversations can also use tools from remote MCP servers and combine client-side and server-side tools.

You can now speak with agents while they work, ask about changes, check progress, and hear spoken updates. Voice calls are available in the Agents Window, IDE, web, and iOS, with multiple voices and zero data retention when Privacy Mode is enabled.

Cloud agents can now create, deploy, and publish web apps with databases, authentication, and hosting. Teams can share these apps securely with coworkers, while built-in checks help prevent exposed secrets and credentials.

The API now accepts additional legacy Office and OpenDocument files, and tools can return documents, audio, and video alongside images and PDFs. This makes it easier to build workflows that analyze and exchange richer files.

The new model supports video generation up to 15 seconds, conversational video editing, and better consistency across up to six reference images. Paid-tier rate limits have also doubled.

A beta feature lets Pro, Max, and Ultra subscribers in the US connect bank, credit card, loan, and investment accounts. Users can view balances and spending in one place and ask questions about their finances.

The new Omni 1.2 Flash can create and edit videos up to 15 seconds long, keep characters and products more consistent, and apply changes through follow-up conversation instead of starting over. It is generally available through the Gemini API and business products, with rollout to several Google apps.

You can upload a photo of an app mock-up, habit tracker, or game board and ask Canvas to build it as an interactive tool. Gemini creates a preview you can test and refine by changing colors, fonts, or features.

The new model is designed for cybersecurity agents that can help uncover and address critical vulnerabilities. It is aimed at more advanced defensive security workflows.

GEO Hub now lets teams compare AI search visibility across nine regions, see what caused score changes, and use GEO findings with Task Agents, the API, and MCP. Business customers can use these updates to understand regional performance and turn recommendations into on-brand content and workflows.