Gemini 3 Pro lands in Gemini and Google Search
Gemini 3 Pro is Google’s new flagship AI model, which the company is releasing to everyone through the Gemini app. Google subscribers can also use the model in AI Mode in Search, with a wider release to everyone in the US coming soon. In the coming weeks, AI Mode and AI Overviews will start “intelligently route your most challenging questions” to the new model while keeping simpler tasks to the older models.Google describes Gemini 3 Pro as a “natively multimodal” model, which can process text, images, video, audio, and code simultaneously, rather than separately. That means you can mix different sources of information and ask the model to synthesize it. For example, you can mix research papers, video lectures, and tutorials, and the model can generate code for flashcards or other interactive learning materials.
Google says Gemini 3 starts a new era of intelligence | Video – Google
In its announcement, Google subtly jabbed at OpenAI, saying that Gemini 3 Pro responses with less flattery, which is a well-known issue with ChatGPT.
Its responses are smart, concise and direct, trading cliché and flattery for genuine insight — telling you what you need to hear, not just what you want to hear.
Google, November 2025
Generative user interface experiences in AI Mode
AI Mode can now generate visual layouts fro the responses of your questions. | Image credit – Google
Gemini 3’s new capabilities unlock various improvements to Google’s apps. For example, you’ll be able to build more comprehensive programs inside the Canvas workspace in the Gemini app. For example, you can let the model recognize handwritten recipes in several languages and build a shareable, multi-language family cookbook.Gemini 3 also enables what Google calls generative user interface experiences. When you look for information in AI Mode, Gemini 3 Pro will create visual layouts for the responses to your questions. Those visualizations feature interactive tools and simulations generated on the fly based on the user’s query.
On top of the benchmarks
Gemini 3 Pro is now available in the Gemini app. | Image credit – PhoneArena
In the coming weeks, Google is also going to launch a more research-intensive version of the model, called Gemini 3 Deep Think. That version will outperform Gemini 3 Pro, which is already leading some independent benchmarks.
Gemini 3 Pro reached 37.5 on the Humanity’s Last Exam benchmark, which is meant to capture general reasoning and expertise. That benchmark was previously led by OpenAI’s GPT-5 Pro with a score of 31.64. Gemini 3 Pro is also leading the leaderboard on LMArena, only hours after xAI’s Grok 4.1 ousted Gemini 2.5 from the top position, which Google’s previous model held for months.
Google’s CEO, Sundar Pichai, says that the Gemini app has more than 650 million monthly active users, and AI Overviews are used by 2 billion people every month. He added that more than 70% of Google’s cloud clients use AI, and over 13 million software developers have used its models as part of their workflows.
Alongside the new model, the company also launched a Gemini-powered coding interface called Google Antigravity. Similar to other agentic IDEs, the app combines a prompt window with a command-line interface and a browser window. The app is available in free preview and is compatible with macOS, Windows, and Linux.
You can’t escape Gemini
While Gemini 3 sounds impressive, I think Google is a bit too aggressive with how it pushes AI to users. AI Overview takes the top part of the Search page, and AI Mode is the leftmost option, positioned before all results, images, and other tabs. That makes it impossible to miss, which could bother everyone who isn’t as excited about AI as Google.
