Gemini 3: Google’s AI Builds Visual Interfaces and Acts Like an Agent

What Gemini 3 brings

Google has unveiled Gemini 3, a significant upgrade to its flagship multimodal model. The company says the new model improves reasoning, expands multimodal capabilities across voice, text, and images, and can operate more like an agent that performs multi-step tasks.

Generative interfaces and 'vibe-coding'

Unlike previous releases that usually return a block of plain text unless given explicit formatting instructions, Gemini 3 introduces what Google calls generative interfaces. These interfaces let the model decide which output format best fits a user prompt and assemble visual layouts and interactive elements on its own. For example, asking for travel recommendations could produce a website-like view inside the app with modules, images, follow-up prompts such as 'How many days are you traveling?' or 'What kinds of activities do you enjoy?', and clickable options that anticipate next steps.

When the model judges that a visual explanation will be more effective, it can generate diagrams or simple animations automatically. Josh Woodward, VP of Google Labs, Gemini, and AI Studio, describes the effect: 'Visual layout generates an immersive, magazine-style view complete with photos and modules. These elements don’t just look good but invite your input to further tailor the results.'

Gemini Agent: handling multi-step tasks

Google is also introducing Gemini Agent, an experimental capability that manages multi-step workflows inside the app. Once users grant access to services like Google Calendar, Gmail, and Reminders, the agent can perform actions such as organizing an inbox or managing schedules. It breaks tasks into discrete steps, displays progress in real time, and pauses for user approval before continuing, positioning itself as a move toward 'a true generalist agent.' The feature will be available on the web for Google AI Ultra subscribers in the US starting November 18.

Deeper integration with Google products

Gemini 3 is woven more tightly into Google's ecosystem. In Search, select Google AI Pro and Ultra subscribers can switch to Gemini 3 Pro, a reasoning-focused variant that delivers deeper, more thorough AI summaries based on the model's reasoning capabilities rather than the existing AI Mode.

For shopping, Gemini will draw on Google's Shopping Graph, which Google says contains over 50 billion product listings, to generate interactive recommendation guides. Users can request shopping help or search shopping phrases, and the model will assemble Wirecutter-style product guides with prices and details without redirecting to external sites.

Tools and opportunities for developers

Google also pushed forward with single-prompt software generation and introduced Google Antigravity, a development platform intended as an all-in-one workspace where code, tools, and workflows can be created and managed from a single prompt.

Industry voices note practical improvements. Derek Nee, CEO of Flowith, an agentic AI application, told MIT Technology Review that Gemini 3 Pro addresses gaps in earlier models, including stronger visual understanding, better code generation, and improved performance on lengthy tasks. 'Given its speed and cost advantages, we’re integrating the new model into our product,' he said, while noting the need for deeper testing to understand its full potential.