So, what else is Gemini up to? Read it here.
1. One Gemini and Skills hub
Personalizing Gemini has become much easier via the new "Customize" button in the sidebar of the web, mobile apps, and desktop applications.
- Through this centralized dashboard, you can manage all your preferences at a glance: Connect third-party services (such as Adobe, OpenTable, or Instacart) and Google apps, allowing Gemini to combine actions across different platforms without switching windows. Set how Gemini should phrase things by default, which formatting rules to follow, and what context the assistant should automatically include.
- Farewell Gems, hello ‘Skills’. Google is saying goodbye to the familiar Gemini Gems and gradually replacing them with Skills. While Gems often felt like isolated mini-bots that you had to navigate to separately, Skills are much more flexible and deeply integrated into the standard chat. You can easily summon a specific skill in any conversation with a /-command (coming soon @). It will also be possible to activate multiple skills simultaneously for a single focused task. Existing Gems will be converted to Skills step by step, so that stored instructions and your knowledge base are neatly preserved.
2. The new heavyweights: Gemini 4 Argon and Carbon
The fourth generation of Gemini models introduces two new flagship models: Gemini 4 Argon and the even more powerful Gemini 4 Carbon. Heavy hitters, do (not) try this at home!
-
Gemini 4 Argon: The workhorse for demanding software engineering, scientific problems, and complex document analysis. Thanks to extensive long-decode-capabilities, Argon can process large reasoning steps and enormous codebases in one continuous sequence.
-
Gemini 4 Carbon: The absolute flagship built on top of Argon. Carbon focuses specifically on sustained agency: long-running, autonomous workflows. The model is designed to compute for hours or even days, independently detect errors in complex code, and automatically course-correct stalled agent processes.
3. Gemini Enterprise transforms into ‘Gemini Agent’
In the business sphere, Google is saying goodbye to the current name Gemini Enterprise in favor of Gemini Agent.
Employees build and manage their own agent(s) in a visual interface without writing a single line of code, thanks to standard connectors (connections to Gmail, Google Drive, Microsoft 365, Jira, Salesforce, and Slack are activated via standard authorization buttons, just like connecting a plugin). You can even go further and implement custom code and containers (BYOC). When you want the agent to consult its own Python scripts or an internal SQL database, Data Warehouse, or internal applications, you set up a Google Cloud Project for this in the Google Cloud Console.
The agent can autonomously schedule appointments, bundle market analyses, which can then be directly translated into a presentation, etc. Within Workspace, your AI colleague will even get its own entity in the company directory, allowing you to simply tag the agent in chats or documents. Stay tuned!
4. Immersive explorations: interactive and visual learning
In the Gemini app and Google Workspace for Education, the immersive explorations feature is being rolled out. Instead of textual explanations, users are presented with a dynamic, visual experience to explore complex topics.
Users can interactively "zoom in" on complex subject matter, such as the anatomy of a jet engine, the formation of a lightning strike, or historical trade routes. By tapping on elements in the visualizations, Gemini reveals deeper layers, interconnections, and scientific data, making it a powerful tool for both students and visual thinkers alike.
And there you have it, et voilà, that's it. We're back on track!
