Google Unveils Gemini 3: A “New Era” of Reasoning and Agentic AI
Google has officially released Gemini 3, its newest generation of AI models. Moving beyond simple text generation, the release focuses on “Deep Think” reasoning, autonomous “agentic” workflows, and a reimagined user experience in the Gemini app featuring “Generative Interfaces.”
What is Gemini 3?
Gemini 3 is a multimodal AI family designed to “read the room” and handle complex, multi-step tasks. The lineup is split into primary versions serving different needs.
The core model, Gemini 3 Pro, is available immediately and acts as an “AI powerhouse,” delivering sharper reasoning and more concise, helpful responses. Users can access this model by selecting “Thinking” from the model selector in the app. It introduces “Vibe Coding” and Generative Interfaces, allowing users to generate rich, interactive experiences from natural language.
In the Gemini app, this manifests as Visual Layouts, which offer magazine-style views for tasks like travel planning, and Dynamic Views, which create custom interactive UIs in real-time, such as a clickable Van Gogh gallery.
For more specialised tasks, Google has introduced Gemini 3 “Deep Think.” This advanced reasoning mode is designed for complex problem-solving and works by “peeling apart overlapping layers” of a problem, taking more time to process intent and context. It has achieved top scores on benchmarks like “Humanity’s Last Exam.”
Additionally, a new experimental tool called Gemini Agent is rolling out to Google AI Ultra subscribers in the U.S. This feature handles complex, multi-step tasks directly within the Gemini interface, such as organising inboxes, prioritising to-dos, and executing real-world tasks like researching and booking a rental car under a specific budget. The agent is designed with a human-in-the-loop approach, seeking confirmation before taking critical actions like making purchases or sending messages.
New Developer Platform: “Antigravity”
Alongside consumer updates, Google launched Google Antigravity, an agentic platform where AI is treated as an active partner. It allows developers to offload end-to-end software tasks, enabling the AI to use terminals and browsers to build and test applications autonomously.
Updates to the Gemini App Experience
Coinciding with the model launch, the Gemini app has received a major redesign to offer a cleaner, more modern look designed to make starting chats easier. A new folder system called “My Stuff” has been added, allowing users to easily find created images, videos, and reports.
Furthermore, the app now features improved shopping integration utilising Google’s Shopping Graph, which leverages over 50 billion product listings to provide better comparison tables and pricing data.
Availability Guide
Google is deploying Gemini 3 immediately across its ecosystem. Gemini 3 Pro is rolling out now globally in the Gemini App and is also powering the new “AI Mode” in Google Search. The experimental Gemini Agent is available on the web today, specifically for Google AI Ultra subscribers.
While the Pro model and Agent are available now, the specialised Gemini 3 Deep Think mode will be released to Google AI Ultra subscribers in the coming weeks, following the completion of final safety evaluations.
