TLDR
- Google launched Gemini 3, its latest AI model, with immediate integration into revenue-generating products including its search engine on day one
- Gemini Agent can now handle multi-step tasks like organizing inboxes and booking travel arrangements without human input
- Paying Google Pro and Ultra subscribers get access to Gemini 3 through AI Mode’s new “Thinking” option in search
- The company introduced Antigravity, a software development platform where AI agents autonomously plan and execute coding tasks
- Gemini 3 scored 37.5% on Humanity’s Last Exam without tool use, demonstrating improved performance in math and science problem-solving
Google unveiled Gemini 3 on Tuesday, marking the first time the company embedded its newest AI model directly into its search engine from launch. The move represents a shift in strategy as Alphabet focuses on monetizing its AI investments while Wall Street watches for signs of return on the technology.
CEO Sundar Pichai called Gemini 3 “our most intelligent model” in a company blog post. The launch comes 11 months after Gemini 2’s release.
Gemini 3 leads on several industry leaderboards that measure AI model performance. However, the AI race has shifted from benchmark comparisons to profit-generating applications of the technology.
Alphabet’s stock has gained this year largely due to financial success from AI offerings in its cloud computing division. Google emphasized that Gemini 3 already powers multiple revenue-generating consumer and enterprise products at launch, unlike previous releases.
“We think Gemini has set quite a new pace in terms of both releasing the models, but also getting it to people faster than ever before,” said Koray Kavukcuoglu, Google’s chief AI architect. Past Gemini versions took weeks or months to integrate into Google’s most-used products.
Paying subscribers to Google’s premium AI tiers can access Gemini 3 capabilities through AI Mode. The search feature replaces standard web results with computer-generated answers for complex queries.
New Agent Capabilities Change User Experience
The company debuted Gemini Agent, which completes multi-step tasks such as organizing inboxes or booking travel. The tool moves Google closer to AI chief Demis Hassabis’ vision for a “universal assistant” previously reported as AlphaAssist.
The Gemini app received a redesign to display answers resembling full websites. This presents another challenge to content publishers who depend on web traffic for revenue.
Josh Woodward, vice president in charge of the app, showed reporters how Gemini now generates on-demand interfaces with visual and interactive elements. A query like “create a Van Gogh gallery with life context for each piece” produces an interactive response.
Improved Visual and Coding Functions
Google claims Gemini 3 is its “most factual model” to date. The model scored 37.5% without tool use on Humanity’s Last Exam, a test comprising 2,500 questions across various subjects.
Tulsee Doshi, product lead for Gemini, said the model “doesn’t just process text or images separately.” It understands nuances across different media types to convert information into the most suitable format.
The model’s enhanced coding abilities allow it to create presentations or interactive graphics to explain complex ideas. Kavukcuoglu believes this will benefit students and anyone using AI as a learning tool.
Google also previewed Antigravity for business customers. The software development platform lets AI agents plan and execute coding tasks autonomously.
The platform demonstrated building an interactive flight-tracking app from a single prompt. The agent creates progress reports during development and a final walkthrough report.
US users paying for Google’s Pro or Ultra Gemini tiers see a new “Thinking” option in Google Search’s AI Mode. Google plans to make Gemini 3 in search available to all users soon.
Robbie Stein, VP of Search, said users have been asking more complex questions as Google rolled out new AI capabilities. Gemini 3 breaks down queries into more pieces for better search performance.
The model can build visualizations and interactive graphics directly on the AI Mode search page. DeepMind CEO Demis Hassabis described the update as going from “simply reading text and images to reading the room.”
Gemini Agent will interact with various Google apps, including managing Google Calendar or rearranging Gmail inbox. Users can ask the agent to research upcoming holidays using email information to find suitable rental cars with minimal human input.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
