Google has introduced Google Gemini 3 Pro, its newest flagship AI model. The launch comes just months after Gemini 2, and Google claims this version delivers its strongest intelligence so far. Users can try the model immediately, as it is rolling out to the Gemini app, AI Mode in Search and Google’s developer platforms.
Gemini 3 Pro focuses on better reasoning, faster coding performance and more reliable multi-modal understanding. The update aims to improve everyday answers while giving advanced users access to new agent-like abilities.
Benchmark results show Google Gemini 3 Pro’s improvements
Google highlights two major benchmarks to showcase Gemini 3 Pro’s progress.
On Humanity’s Last Exam, the model reached 37.5% accuracy, which places it above Grok 4 by 12.1 percentage points. Impressively, it reached this score without relying on external tools such as a web search engine.
On LMArena, Gemini 3 Pro reached 1,501 points, moving it to the top of the community leaderboard. Google uses these numbers to argue that the new model offers stronger reasoning than earlier versions.
What changes inside the Gemini app
In the Gemini app, Google Gemini 3 Pro gives shorter, clearer and more structured answers. The model also unlocks a new feature called Gemini Agent.
This tool can now complete tasks for users. For example, instead of suggesting how to manage an inbox, it can carry out the organizing steps directly when granted access to your Google apps.
To access Gemini 3 Pro, users can select “Thinking” in the model picker. All users can try the model, although AI Plus, Pro and Ultra subscribers receive a larger usage allowance before hitting rate limits.
Google Gemini 3 Pro expands into Search and AI Mode
The model also arrives inside Google Search. In AI Mode, it will first appear for AI Pro and Ultra subscribers. Google plans to expand access gradually.
Gemini 3 Pro will also support AI Overviews, especially when handling complex or multi-step queries.
Google is preparing a new routing system that decides when a question should use Gemini 3 Pro instead of another model. Until then, subscribers can switch to “Thinking” in AI Mode to activate it manually.
With Gemini 3 Pro, AI Mode should surface more reliable information. The system now performs more parallel searches and evaluates a wider range of sources. According to Google, the improved reasoning lets the model discover content that older versions often overlooked.
Stronger multi-modal skills also allow AI Mode to create interactive elements. When someone researches mortgage loans, the response may include a loan calculator instead of a plain text explanation.
Tools for developers and enterprises
Developers and enterprise users can access Google Gemini 3 Pro in the Gemini API, AI Studio and Vertex AI.
Alongside the new model, Google introduced Antigravity, an autonomous coding tool. It can plan tasks, write code and report progress without constant user input.
Google is also releasing Gemini 3 Deep Think, a mode designed for advanced reasoning. Safety testers will access it first, followed by AI Ultra subscribers.
How Google Gemini 3 Pro changes the overall experience
Gemini 3 Pro shows how quickly Google is moving toward more agent-like AI tools. In the Gemini app, the model provides cleaner answers and performs tasks directly. In Search, it supports deeper web exploration and more dynamic responses. Developers gain a stronger engine for building AI-powered applications.
By launching Google Gemini 3 Pro less than a year after Gemini 2, Google signals that rapid iteration is now the norm. Users can expect frequent upgrades, while the wider AI landscape becomes more competitive.
Read also
Join the discussion in our Facebook community.